CN111126629A - Model generation method, system, device and medium for identifying brushing behavior - Google Patents
Model generation method, system, device and medium for identifying brushing behavior Download PDFInfo
- Publication number
- CN111126629A CN111126629A CN201911356516.9A CN201911356516A CN111126629A CN 111126629 A CN111126629 A CN 111126629A CN 201911356516 A CN201911356516 A CN 201911356516A CN 111126629 A CN111126629 A CN 111126629A
- Authority
- CN
- China
- Prior art keywords
- order
- orders
- sample data
- negative sample
- brushing
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- 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
- G06F18/211—Selection of the most significant subset of features
- G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
-
- 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
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0633—Lists, e.g. purchase orders, compilation or processing
- G06Q30/0635—Processing of requisition or of purchase orders
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Finance (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Accounting & Taxation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a model generation method, a single-line-brushing identification method, a system, equipment and a medium, wherein the model generation method comprises the following steps: acquiring a plurality of historical orders, and constructing model data based on the historical orders; constructing basic information characteristics based on the model data; constructing context information characteristics based on the trend of the single brushing behavior and the model data; and training by adopting an ensemble learning method based on the basic information features and the context information features to obtain the brushing line as an identification model. The method and the device can identify the manually unknown brushing list behavior, and further improve the recall rate of the brushing list identification.
Description
Technical Field
The invention belongs to the field of artificial intelligence, and particularly relates to a model generation method, a single-line-brushing identification method, a system, equipment and a medium.
Background
In recent years, along with increasingly specialized bill-swiping means, building a bill-swiping identification system in various application scenes of electronic commerce can effectively reduce the benefit loss of the bill-swiping behavior to users and platforms.
The method for constructing the bill-swiping identification system in various application scenes such as tourism scenes needs to consider various complex factors. The current order brushing identification is mainly based on a rule system, and different rules are formulated through the field knowledge and experience of human experts for identifying abnormal orders with the order brushing. However, the rule system only considers the characteristic of the single brushing line obtained by statistics in history, and cannot identify the manually unknown single brushing behavior; in addition, with the continuous promotion and specialization of the one-hand brushing method, the existing rule system needs to be continuously updated and iterated, and more missed grabbing and mistaken grabbing can occur in the process.
Disclosure of Invention
The invention aims to overcome the defects that different rules are formulated for identifying abnormal orders with a single-line-swiping behavior and the manually unknown single-line-swiping behavior cannot be identified through the field knowledge and experience of manual experts based on a rule system in the prior art, and provides a model generation method, a single-line-swiping behavior identification method, a system, equipment and a medium.
The invention solves the technical problems through the following technical scheme:
the invention provides a method for generating a single-row-brushing recognition model, which comprises the following steps:
acquiring a plurality of historical orders, and constructing model data based on the historical orders;
constructing basic information characteristics based on the model data;
constructing context information characteristics based on the trend of the single brushing behavior and the model data;
and training by adopting an ensemble learning method based on the basic information features and the context information features to obtain the brushing line as an identification model.
In the scheme, the brushing single row is obtained as the recognition model by constructing model data and context information characteristics and adopting an integrated learning method for training, so that an accurate recognition model is provided for realizing the recognition of the brushing single row.
Preferably, the step of constructing the context information feature based on the tendency of the swipe behavior and the model data includes:
acquiring one of a plurality of historical orders as a current order at each time, and acquiring a product and a supplier corresponding to the current order;
acquiring N historical orders of the product, which are adjacent to the current order, as first orders to be screened;
acquiring N historical orders of the supplier, which are adjacent to the current order, as second orders to be screened;
and calculating the order proportion and/or the order number meeting the preset screening condition based on the first order to be screened and the second order to be screened, and taking the order proportion and/or the order number as the context information characteristic.
In the scheme, the step of constructing the context information characteristics based on the tendency of the single-line-brushing behavior and the model data is provided, the problem that the tendency of the single-line-brushing behavior cannot be reflected due to the fact that the basic information characteristics of all dimensions can reflect the information of the current order to a certain extent is solved, and the accuracy of training for identifying the model by single-line-brushing is improved.
Preferably, the contextual information features include at least one of the order number of the same hardware device but different user names, the order proportion of the orders with the same payment type, the order proportion of the orders of similar users, and the order proportion of the orders with abnormal mobile phone numbers and/or certificate numbers.
In this scheme, the four context features provided are only four types divided according to types, but actually, the total number of the context features is greater than four.
Preferably, when the contextual information features include the order proportion occupied by the orders with the same payment type and/or the orders with the same hardware device and different user names, the step of calculating the order proportion occupied by the orders with the same payment type and/or the step of calculating the orders with the same hardware device and different user names includes:
acquiring any one of a plurality of historical orders as a current order, and acquiring a product and a supplier corresponding to the current order;
grouping a plurality of historical orders according to the same products and the same suppliers respectively to obtain a first order group and a second order group;
screening N adjacent orders which are the same as the products and suppliers of the current order from the first order group and the second order group to obtain a third order group;
when the context information features include the order proportion occupied by the orders with the same payment type, the step of calculating the order proportion occupied by the orders with the same payment type further includes:
screening n orders with the same order payment mode as the current order in the third order group;
calculating the order proportion occupied by the orders with the same payment type according to a context characteristic calculation formula;
the context feature calculation formula is as follows:
wherein R issamePayRepresenting the order proportion occupied by the orders with the same payment type;
when the context information feature comprises the same amount of orders of the hardware devices but different user names, the step of calculating the amount of orders of the same hardware devices but different user names further comprises:
screening out an order identical to the hardware equipment of the current order from the third order group to obtain a fifth order group;
and counting the number of the user names in the fifth order group as the number of orders with the same hardware equipment but different user names.
Preferably, the step of obtaining a plurality of historical orders and constructing model data based on the historical orders comprises:
acquiring a plurality of historical orders of various channels;
labeling each historical order to obtain first sample data, wherein the labeling result comprises a brushing list and a non-brushing list; the first sample data with the labeling result of the list brushing is first positive sample data, and the first sample data with the labeling result of the list brushing is first negative sample data;
carrying out negative sample expansion based on the prior knowledge to obtain second negative sample data;
merging the second negative sample data and the first negative sample data to obtain third negative sample data;
and sampling the first positive sample data and the third negative sample data according to a preset positive-negative sample proportion to obtain the model data.
In the scheme, the proportion of the positive and negative samples is limited, the problem of large scale of the expanded negative samples is solved, and better model data is obtained.
Preferably, the step of performing negative sample expansion based on the prior knowledge to obtain second negative sample data comprises:
acquiring orders with low scores and complaints at the same time from a basic order data set as second negative sample data;
the step of sampling the first positive sample data and the third negative sample data according to a preset positive-negative sample ratio to obtain the model data includes:
limiting the proportion of the first positive sample data and the third negative sample data according to a proportion limiting formula;
the formula of the ratio limit is as follows:
wherein N ispos,Nneg,NaddRespectively representing instances of the first positive sample data, the first negative sampleThe case number of data, the case number of the second negative sample data, neg2posRate, which represents a ratio of the case number of the first positive sample data to the case number of the third negative sample data, and negSampleRate, which represents a ratio of the first negative sample data;
the model data comprises the first positive sample data and the third negative sample data.
Preferably, the step of training the brushing row to obtain the recognition model based on the basic information features and the context information features by using an ensemble learning method includes:
and training by adopting an XGboost (Extreme Gradient Boosting) model in an ensemble learning method based on the basic information features and the context information features to obtain the brush line as an identification model.
In this scheme, the characteristic performance based on the order of brushing is comparatively complicated, and different brush one-way sections have different action preferences, have adopted XGboost model training to obtain stable and each side performance better brush the single file for the recognition model.
The invention also provides a single-line-brushing-action identification method, which comprises the following steps:
acquiring order data to be identified;
inputting the order data to be identified into a single-file-brushing-line identification model for identification so as to obtain a single-file-brushing-line identification result;
the brush single line is generated for the identification model by using the generation method of the brush single line for the identification model.
In the scheme, the generation method of using the single brushing row as the recognition model is adopted to generate the single brushing row as the recognition model, so that the artificially unknown single brushing behavior can be recognized, and the recall rate of the single brushing recognition is further improved.
The invention also provides a generation system of the recognition model of the single-line swiping behavior, which comprises the following steps:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring a plurality of historical orders and constructing model data based on the historical orders;
a first construction module for constructing basic information features based on the model data;
the second construction module is used for constructing context information characteristics based on the trend of the single brushing behavior and the model data;
and the training module is used for training by adopting an ensemble learning method based on the basic information characteristics and the context information characteristics to obtain the brushing row as a recognition model.
Preferably, the second building block comprises:
the first obtaining unit is used for obtaining one from a plurality of historical orders as a current order at a time and obtaining a product and a supplier corresponding to the current order;
a second obtaining unit, configured to obtain N historical orders of the product that are adjacent to the current order and serve as first orders to be screened;
a third obtaining unit, configured to obtain N historical orders of the supplier that are adjacent to the current order and serve as a second order to be screened;
and the screening unit is used for calculating the order proportion and/or the order number meeting the preset screening condition based on the first order to be screened and the second order to be screened, and taking the order proportion and/or the order number as the context information characteristic.
Preferably, the contextual information features include at least one of the order number of the same hardware device but different user names, the order proportion of the orders with the same payment type, the order proportion of the orders of similar users, and the order proportion of the orders with abnormal mobile phone numbers and/or certificate numbers.
Preferably, when the contextual information features include the order proportion occupied by the orders with the same payment type and/or the number of orders with the same hardware device but different user names, the screening unit includes a corresponding calculating subunit of the order proportion occupied by the orders with the same payment type and/or a corresponding calculating subunit of the number of orders with the same hardware device but different user names; the screening unit further comprises a preprocessing subunit;
the preprocessing subunit is configured to acquire any one of the plurality of historical orders as a current order, and acquire a product and a supplier corresponding to the current order; grouping a plurality of historical orders according to the same products and the same suppliers respectively to obtain a first order group and a second order group; screening N adjacent orders which are the same as the products and suppliers of the current order from the first order group and the second order group to obtain a third order group;
the calculation subunit of the order proportion occupied by the orders with the same payment type is used for screening n orders with the same order payment mode as the current order in the third order group; calculating the order proportion occupied by the orders with the same payment type according to a context characteristic calculation formula;
the context feature calculation formula is as follows:
wherein R issamePayRepresenting the order proportion occupied by the orders with the same payment type;
the calculation subunit of the same hardware equipment but with different orders of the user name is used for screening out the order which is the same as the hardware equipment of the current order from the third order group to obtain a fifth order group; and counting the number of user names in the fifth order group as the number of orders with the same hardware equipment but different user names.
Preferably, the first building block comprises:
the fourth acquisition unit is used for acquiring a plurality of historical orders of various channels;
the labeling unit is used for labeling each historical order to obtain first sample data, and the labeling result comprises a brushing list and a non-brushing list; the first sample data with the labeling result of the list brushing is first positive sample data, and the first sample data with the labeling result of the list brushing is first negative sample data;
the expansion unit is used for carrying out negative sample expansion based on the prior knowledge to obtain second negative sample data;
the merging unit is used for merging the second negative sample data and the first negative sample data to obtain third negative sample data;
the sampling unit is used for sampling the first positive sample data and the third negative sample data according to a preset positive-negative sample proportion to obtain the model data.
Preferably, the expansion unit is further configured to obtain an order with a low score and a complaint from a basic order data set as the second negative sample data;
the sampling unit is further configured to limit a proportion of the first positive sample data and the third negative sample data according to a formula for proportional limitation;
the formula of the ratio limit is as follows:
wherein N ispos,Nneg,NaddRespectively representing the number of instances of the first positive sample data, the number of instances of the first negative sample data, the number of instances of the second negative sample data, neg2posRate representing the ratio of the number of instances of the first positive sample data to the number of instances of the third negative sample data, and negSampleRate representing the ratio of the first negative sample data;
the model data comprises the first positive sample data and the third negative sample data.
Preferably, the training module is configured to train, based on the basic information features and the context information features, by using an XGBoost model in an ensemble learning method to obtain the swiped line as an identification model.
The invention also provides a system for identifying the brushing single line, which comprises:
the second acquisition module is used for acquiring order data to be identified;
the identification module is used for inputting the order data to be identified into the single-row-brushing identification model for identification so as to obtain a single-row-brushing identification result;
the brushing row is generated for the recognition model by using the generating system of the brushing row for the recognition model.
The invention further provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to realize the generation method of the brushing single line as the identification model or the brushing single line as the identification method.
The present invention also provides a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method for generating a swipe line as an identification model or the steps of a swipe line as an identification method as described above.
The positive progress effects of the invention are as follows: the invention provides a method for generating a model, a method, a system, equipment and a medium for identifying a single-line-of-brush, wherein a single-line-of-brush is generated as an identification model through the construction of model data, the construction of model characteristics and the construction and training of the model, and further, the single-line-of-brush can be identified by utilizing the model. Compared with a rule-based system, different rules are formulated through the field knowledge and experience of artificial experts for identifying the existence of the brush single-row behavior.
Drawings
Fig. 1 is a flowchart of a method for generating a brush line recognition model according to embodiment 1 of the present invention.
Fig. 2 is a flowchart of step S103 in embodiment 1 of the present invention.
Fig. 3 is a flowchart of the steps of calculating the order proportion of the orders with the same payment type according to embodiment 1 of the present invention.
Fig. 4 is a flowchart of the steps of calculating the same amount of orders of the same hardware device but different user names according to embodiment 1 of the present invention.
Fig. 5 is a flowchart of step S102 in embodiment 1 of the present invention.
Fig. 6 is a flowchart of a method for identifying a brush line according to embodiment 2 of the present invention.
Fig. 7 is a block diagram of a system for generating a brush line recognition model according to embodiment 3 of the present invention.
Fig. 8 is a schematic structural diagram of a second building block in embodiment 3 of the present invention.
Fig. 9 is a schematic structural diagram of a first building block in embodiment 3 of the present invention.
Fig. 10 is a block diagram of a system for identifying a brush line in embodiment 4 of the present invention.
Fig. 11 is a schematic structural diagram of an electronic device according to embodiment 5 of the present invention.
Detailed Description
The invention is further illustrated by the following examples, which are not intended to limit the scope of the invention.
Example 1
As shown in fig. 1, the present embodiment discloses a method for generating a brushing line as an identification model, which includes the following steps:
s101, acquiring a plurality of historical orders, and constructing model data based on the historical orders;
s102, constructing basic information characteristics based on the model data;
in this embodiment, the basic information features include a user dimension feature, an order basic attribute dimension feature, and a product dimension feature. The user dimension characteristics include: the behavior data of the user in the last 15 days is compared with the behavior difference of the user who swipes the order and the user who does not swipe the order, the behavior data of the user who swipes the order in the last 15 days is different from the distribution of the behavior data of the user who does not swipe the order, and whether the user has abnormality before placing the order can be reflected through the behavior data characteristics; the time interval between user registration, order placement and last login, etc. The order basic attribute dimension characteristics comprise: product form, product type, source of order; the number of people going out of the order; sales, cost, gross profit, and derived features calculated based on other existing features, such as average daily sales per person for the order; the travel days, the preset days in advance, the month of the order, whether the order has an air ticket, whether the order has insurance and the like. The product dimensional characteristics include: the number of users ordering and browsing the order products on different days; conversion of the ordered product on the current day, average daily conversion, relative fluctuation between the two.
S103, constructing context information characteristics based on the trend of the order brushing behavior and the model data;
the basic information characteristics can reflect the information of the current order to a certain extent, but the trend of the order-swiping behavior cannot be reflected. A typical brush line is known to be a feature that typically accompanies batch runs, i.e., there are similarly behaving batch orders in the vicinity of the brush order. For example, only the id of the order user is taken as a feature input model, and the id is weakly associated with the brush line as a feature, so that useful information cannot be provided for the model training process. And calculating the proportion of similar user ids in adjacent orders placed by the same product of the order based on the trend of the order brushing behavior, wherein if the proportion is close to 1, the situation that batch registered users place orders for the product exist in a period of time is shown. Therefore, the suspicion that the current order is a scrub order increases. The embodiment develops the context characteristics for the order based on the characteristics of the order-swiping behavior, namely the tendency of the order-swiping behavior.
And S104, training by adopting an ensemble learning method based on the basic information features and the context information features to obtain the brushing line as an identification model.
In this embodiment, the brushed line is obtained as the recognition model by using XGBoost model training in the ensemble learning method based on the basic information features and the context information features. In the embodiment, a tour scene is used as an example, whether an order is swiped for a single line is judged in the tour scene, and a supervised learning algorithm is adopted in order to learn a stable model which is good in all aspects. The characteristic expression of the order form brushing under the scene of travel is complex, different order brushing sections have different behavior preferences, the model actually trained is a weak supervision model, and the order form brushing under a certain preference has better recognition capability. The ensemble learning is expected to obtain a more comprehensive strong supervision model by combining a plurality of weak supervision learners, and is suitable for detecting the single brushing line in a tourism scene. The XGboost model in ensemble learning is adopted to carry out two classification tasks, and the algorithm idea is to generate a tree by continuously adding tree models and continuously carrying out feature splitting; every new tree is generated, and in fact a new function is learned to fit the residual of the last round of prediction. And obtaining K subtrees after training is finished, wherein each sample falls to a corresponding leaf node in each tree, each leaf node corresponds to a score, and the final predicted value of the sample is the sum of the scores of the leaf nodes of each corresponding tree. The objective function for the K-th round is expressed as follows:
wherein, ObjKRepresenting the objective function of the K-th round, yiThe (i) th real value is represented,
represents the ith prediction value in the K round, L () represents a loss function, fKRepresents the kth sub-tree, Ω () represents the sub-tree complexity, constant represents a constant.
In the XGBoost, a gradient is obtained by performing second-order taylor expansion on a loss function, and a specific calculation formula is as follows:
wherein the content of the first and second substances,
denotes the ith prediction value, f, in the K-1 th roundK(xi) Representing the prediction value of the kth sub-tree for the ith sample.
The detailed process of the algorithm is the prior art and is not described herein.
As shown in fig. 2, in the present embodiment, step S103 includes the following steps:
step S1031, obtaining an order from a plurality of historical orders as a current order, and obtaining a product and a supplier corresponding to the current order;
step S1032, acquiring N historical orders of the product, which are adjacent to the current order, as first orders to be screened;
step S1033, obtaining N historical orders of the supplier adjacent to the current order as second orders to be screened;
step S1034, calculating an order proportion and an order number meeting a preset screening condition based on the first order to be screened and the second order to be screened, and using the order proportion and the order number as the context information feature. In other alternative embodiments, the order proportion and the order number may comprise only one.
In this embodiment, the context characteristics of each order in the historical orders are sequentially constructed by using the above steps, and the context information characteristics include the order number with the same hardware equipment but different user names, the order proportion occupied by the orders with the same payment type, the order proportion occupied by the orders with similar users, and the order proportion occupied by the orders with abnormal mobile phone numbers and/or certificate numbers. In alternative embodiments, only one or more of the above-described contextual characteristics may be included.
As shown in fig. 3, in this embodiment, the step of calculating the ratio of orders with the same payment type includes:
s1051, obtaining any one of a plurality of historical orders as a current order, and obtaining a product and a supplier corresponding to the current order;
s1052, grouping the plurality of historical orders according to the same products and the same suppliers respectively to obtain a first order group and a second order group;
s1053, screening N adjacent orders which are the same as the products and suppliers of the current order from the first order group and the second order group to obtain a third order group;
s1054, screening n orders with the same order payment mode as the current order in the third order group;
calculating the order proportion occupied by the orders with the same payment type according to a context characteristic calculation formula;
the context feature calculation formula is as follows:
wherein R issamePayAnd the order proportion of the orders with the same payment type is represented.
As shown in fig. 4, in this embodiment, the step of calculating the amount of orders with the same hardware device but different user names includes:
s1061, acquiring any one of the historical orders as a current order, and acquiring a product and a supplier corresponding to the current order;
s1062, grouping the historical orders according to the same products and the same suppliers respectively to obtain a first order group and a second order group;
s1063, screening out adjacent orders which are the same as the products and suppliers of the current order from the first order group and the second order group to obtain a third order group;
s1064, screening out an order identical to the hardware equipment of the current order from the third order group to obtain a fifth order group;
and S1065, counting the number of the user names in the fifth order group as the number of orders with the same hardware equipment but different user names.
As shown in fig. 5, in the present embodiment, step S102 includes the following steps:
s1021, acquiring a plurality of historical orders of various channels;
in this embodiment, the historical orders may be obtained from multiple channels, such as orders that are manually confirmed as being swiped after being reported by a service, orders that are successfully filed by a supplier after being reported, and orders that are manually sampled and determined, and the historical order data obtained from different channels are combined. In the process of merging data, if the same order appears in a plurality of data tables at the same time, a record with the highest priority is reserved according to the priority specified in advance.
S1022, labeling each historical order to obtain first sample data, wherein the labeling result comprises a bill brushing result and a non-bill brushing result; the first sample data with the labeling result of the list brushing is first positive sample data, and the first sample data with the labeling result of the list brushing is first negative sample data;
s1023, carrying out negative sample expansion based on prior knowledge to obtain second negative sample data;
in this embodiment, the order with the low score and the complaint at the same time is obtained from the basic order data set as the second negative sample data;
s1024, merging the second negative sample data and the first negative sample data to obtain third negative sample data;
and S1025, sampling the first positive sample data and the third negative sample data according to a preset positive-negative sample proportion to obtain model data, wherein the model data comprises the first positive sample data and the third negative sample data.
In this embodiment, a formula for limiting the proportion limits the proportion of the first positive sample data and the third negative sample data;
the formula of the ratio limit is as follows:
wherein N ispos,Nneg,NaddRespectively representing the number of instances of the first positive sample data, the number of instances of the first negative sample data, the number of instances of the second negative sample data, neg2posRate representing the ratio of the number of instances of the first positive sample data to the number of instances of the third negative sample data, and negSampleRate representing the ratio of the first negative sample data;
according to the method for generating the identification model by using the single-file-brushing manner, the single-file-brushing manner is generated as the identification model by constructing the model data, constructing the model characteristics and constructing and training the model, different rules are formulated for identifying the existence of the single-file-brushing manner by comparing the method based on the rule system and the field knowledge and experience of artificial experts, the embodiment can identify the artificial unknown single-file-brushing behavior, and the recall rate of the single-file-brushing identification is further improved.
Example 2
As shown in fig. 6, the present embodiment provides a method for identifying a brush line, including the following steps:
step S201, obtaining order data to be identified;
step S202, inputting the order data to be identified into a single-line-swiping identification model for identification so as to obtain a single-line-swiping identification result;
step S203, the single-row-by-brush recognition model is generated by using the aforementioned method for generating a single-row-by-brush recognition model.
According to the method for identifying the brushing single line, the order data to be identified is acquired and is input into the brushing single line identification model for identification, so that the result that the brushing single line is identified is obtained, the manual unknown brushing single line can be identified, and the recall rate of the brushing single identification is further improved.
Example 3
As shown in fig. 7, the present embodiment provides a generation system of a swipe behavior recognition model, including: the system comprises a first acquisition module 1, a first construction module 2, a second construction module 3 and a training module 4.
The first acquisition module 1 is used for acquiring a plurality of historical orders and constructing model data based on the historical orders;
the first construction module 2 is used for constructing basic information characteristics based on the model data;
the second construction module 3 is used for constructing context information characteristics based on the tendency of the brush line and the model data;
the training module 4 is used for training by adopting an XGboost model in an ensemble learning method based on the basic information features and the context information features to obtain the brushing line as an identification model.
As shown in fig. 8, in the present embodiment, the second building block 3 includes a first acquiring unit 31, a second acquiring unit 32, a third acquiring unit 33, and a screening unit 34.
The first obtaining unit 31 is configured to obtain one of the historical orders as a current order each time, and obtain a product and a supplier corresponding to the current order;
the second obtaining unit 32 is configured to obtain N historical orders of the product adjacent to the current order as a first order to be screened;
the third obtaining unit 33 is configured to obtain N historical orders of the supplier adjacent to the current order as a second order to be screened;
the screening unit 34 is configured to calculate an order proportion and an order number meeting a preset screening condition based on the first order to be screened and the second order to be screened, and use the order proportion and the order number as the contextual information features. In other alternative embodiments, the order proportion and the order number may comprise only one.
In this embodiment, the contextual information features include the number of orders with the same hardware device but different user names, the order proportion occupied by orders with the same payment type, the order proportion occupied by orders with similar users, and the order proportion occupied by orders with abnormal cell phone numbers and/or certificate numbers. In alternative embodiments, only one or more of the foregoing may be included.
In this embodiment, the screening unit 34 includes a preprocessing subunit and a calculating subunit of the order proportion occupied by orders with the same payment type;
the preprocessing subunit is used for acquiring any one of the plurality of historical orders as a current order and acquiring a product and a supplier corresponding to the current order; grouping a plurality of historical orders according to the same products and the same suppliers respectively to obtain a first order group and a second order group; screening N adjacent orders which are the same as the products and suppliers of the current order from the first order group and the second order group to obtain a third order group;
the calculation subunit of the order proportion occupied by the orders with the same payment type is used for screening n orders with the same order payment mode as the current order in the third order group; calculating the order proportion of orders with the same payment type according to a context characteristic calculation formula;
the context feature calculation formula is as follows:
wherein R isasmePayRepresenting the proportion of orders with the same payment type;
in this embodiment, the screening unit 34 further includes a calculating subunit having the same hardware device but different orders of the user name;
the calculation subunit of the same hardware equipment but with different orders of the user name is used for screening out the order which is the same as the hardware equipment of the current order from the third order group to obtain a fifth order group; and counting the number of user names in the fifth order group as the number of orders with the same hardware equipment but different user names.
As shown in fig. 9, in the present embodiment, the first building block 2 includes: a fourth obtaining unit 21, a labeling unit 22, an expansion unit 23, a merging unit 24, and a sampling unit 25.
The fourth obtaining unit 21 is configured to obtain a plurality of historical orders of multiple channels;
the labeling unit 22 is configured to label each historical order to obtain first sample data, where the labeling result includes a brushing list and a non-brushing list; the first sample data with the result of labeling as the bill-swiping is first positive sample data, and the first sample data with the result of labeling as the non-bill-swiping is first negative sample data;
the expansion unit 23 is configured to perform negative sample expansion based on the priori knowledge to obtain second negative sample data; acquiring orders with low scores and complaints simultaneously from the basic order data set as the second negative sample data;
the merging unit 24 is configured to merge the second negative sample data with the first negative sample data to obtain third negative sample data;
the sampling unit 25 is configured to sample the first positive sample data and the third negative sample data according to a preset positive-negative sample ratio to obtain the model data, where the model data includes the first positive sample data and the third negative sample data. Specifically, a formula of scale limitation limits a scale of the first positive sample data and the third negative sample data;
the formula of the ratio limit is as follows:
wherein N ispos,Nneg,NaddRespectively representing the number of instances of the first positive sample data, the number of instances of the first negative sample data, the number of instances of the second negative sample data, neg2posRate representing the ratio of the number of instances of the first positive sample data to the number of instances of the third negative sample data, and negSampleRate representing the ratio of the first negative sample data;
the generation system for the identification model of the single swiping line provided by the embodiment generates a method for identifying the existence of the single swiping line by establishing model data, establishing model characteristics, establishing a model and training, and establishing different rules through the field knowledge experience of an artificial expert on the basis of a rule system.
Example 4
As shown in fig. 10, the present embodiment provides a system for swipes a single line as an identification, including: a second obtaining module 5 and an input module 6.
The second obtaining module 5 is used for obtaining order data to be identified;
the input module 6 is used for inputting the order data to be identified into the single-row-swiping identification model for identification so as to obtain a single-row-swiping identification result;
the swipe row is generated for the recognition model using the generation system of the swipe row for the recognition model of example 3.
According to the system for identifying the brushing single line, the order data to be identified is acquired and is input into the brushing single line identification model for identification, so that the result that the brushing single line is identified is obtained, the manual unknown brushing single line can be identified, and the recall rate of the brushing single identification is further improved.
Example 5
Fig. 11 is a schematic structural diagram of an electronic device according to embodiment 5 of the present invention. The electronic device comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the generation method of the brush line recognition model provided by the embodiment 1 and the generation method of the brush line recognition model provided by the embodiment 2. The electronic device 30 shown in fig. 11 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiment of the present invention.
As shown in fig. 11, the electronic device 30 may be embodied in the form of a general purpose computing device, which may be, for example, a server device. The components of the electronic device 30 may include, but are not limited to: the at least one processor 31, the at least one memory 32, and a bus 33 connecting the various system components (including the memory 32 and the processor 31).
The bus 33 includes a data bus, an address bus, and a control bus.
The memory 32 may include volatile memory, such as Random Access Memory (RAM)321 and/or cache memory 322, and may further include Read Only Memory (ROM) 323.
The processor 31 executes various functional applications and data processing, such as the generation method of the brush line recognition model provided in embodiment 1 and the brush line recognition method provided in embodiment 2 of the present invention, by running the computer program stored in the memory 32.
The electronic device 30 may also communicate with one or more external devices 34 (e.g., keyboard, pointing device, etc.). Such communication may be through input/output (I/O) interfaces 35. Also, model-generating device 30 may also communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet) via network adapter 36. As shown, network adapter 36 communicates with the other modules of model-generating device 30 via bus 33. It should be understood that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the model-generating device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.
It should be noted that although in the above detailed description several units/modules or sub-units/modules of the electronic device are mentioned, such a division is merely exemplary and not mandatory. Indeed, the features and functionality of two or more of the units/modules described above may be embodied in one unit/module according to embodiments of the invention. Conversely, the features and functions of one unit/module described above may be further divided into embodiments by a plurality of units/modules.
Example 6
The present embodiment provides a computer-readable storage medium on which a computer program is stored, the program, when executed by a processor, implementing the method for generating a brush line as an identification model provided in embodiment 1 and the steps of the brush line as an identification method provided in embodiment 2.
More specific examples, among others, that the readable storage medium may employ may include, but are not limited to: a portable disk, a hard disk, random access memory, read only memory, erasable programmable read only memory, optical storage device, magnetic storage device, or any suitable combination of the foregoing.
In a possible implementation manner, the present invention can also be implemented in the form of a program product, which includes program code for causing a terminal device to execute the steps of implementing the method for generating the brush line for recognition model provided in embodiment 1 and the method for generating the brush line for recognition provided in embodiment 2, when the program product is run on the terminal device.
Where program code for carrying out the invention is written in any combination of one or more programming languages, the program code may be executed entirely on the user device, partly on the user device, as a stand-alone software package, partly on the user device and partly on a remote device or entirely on the remote device.
While specific embodiments of the invention have been described above, it will be appreciated by those skilled in the art that this is by way of example only, and that the scope of the invention is defined by the appended claims. Various changes and modifications to these embodiments may be made by those skilled in the art without departing from the spirit and scope of the invention, and these changes and modifications are within the scope of the invention.
Claims (16)
1. A method for generating a brush line as an identification model, comprising:
acquiring a plurality of historical orders, and constructing model data based on the historical orders;
constructing basic information characteristics based on the model data;
constructing context information characteristics based on the trend of the single brushing behavior and the model data;
and training by adopting an ensemble learning method based on the basic information features and the context information features to obtain the brushing line as an identification model.
2. The method of generating a brush line for an identification model according to claim 1,
the step of constructing context information features based on the tendency of the bill-swiping behavior and the model data comprises the following steps:
acquiring one of a plurality of historical orders as a current order at each time, and acquiring a product and a supplier corresponding to the current order;
acquiring N historical orders of the product, which are adjacent to the current order, as first orders to be screened;
acquiring N historical orders of the supplier, which are adjacent to the current order, as second orders to be screened;
and calculating the order proportion and/or the order number meeting the preset screening condition based on the first order to be screened and the second order to be screened, and taking the order proportion and/or the order number as the context information characteristic.
3. The method for generating a swipe row recognition model according to claim 2, wherein the contextual information features include at least one of the number of orders with the same hardware device but different user names, the ratio of orders with the same payment type, the ratio of orders with similar users, and the ratio of orders with abnormal phone numbers and/or certificate numbers.
4. The method for generating a swipe line for identifying a model as claimed in claim 3, wherein when the contextual information features include the order proportion occupied by the orders with the same payment type and/or the orders with the same hardware device but different user names, the step of calculating the order proportion occupied by the orders with the same payment type and/or the step of calculating the orders with the same hardware device but different user names each includes:
acquiring any one of a plurality of historical orders as a current order, and acquiring a product and a supplier corresponding to the current order;
grouping a plurality of historical orders according to the same products and the same suppliers respectively to obtain a first order group and a second order group;
screening N adjacent orders which are the same as the products and suppliers of the current order from the first order group and the second order group to obtain a third order group;
when the context information features include the order proportion occupied by the orders with the same payment type, the step of calculating the order proportion occupied by the orders with the same payment type further includes:
screening n orders with the same order payment mode as the current order in the third order group;
calculating the order proportion occupied by the orders with the same payment type according to a context characteristic calculation formula;
the context feature calculation formula is as follows:
wherein R issamePayRepresenting the order proportion occupied by the orders with the same payment type;
when the context information feature comprises the same amount of orders of the hardware devices but different user names, the step of calculating the amount of orders of the same hardware devices but different user names further comprises:
screening out an order identical to the hardware equipment of the current order from the third order group to obtain a fifth order group;
and counting the number of the user names in the fifth order group as the number of orders with the same hardware equipment but different user names.
5. The method of generating a brush line for identification models according to claim 1, wherein said step of obtaining a plurality of historical orders, and constructing model data based on said historical orders comprises:
acquiring a plurality of historical orders of various channels;
labeling each historical order to obtain first sample data, wherein the labeling result comprises a brushing list and a non-brushing list; the first sample data with the labeling result of the list brushing is first positive sample data, and the first sample data with the labeling result of the list brushing is first negative sample data;
carrying out negative sample expansion based on the prior knowledge to obtain second negative sample data;
merging the second negative sample data and the first negative sample data to obtain third negative sample data;
and sampling the first positive sample data and the third negative sample data according to a preset positive-negative sample proportion to obtain the model data.
6. The method of claim 5, wherein the step of performing negative sample expansion based on the prior knowledge to obtain the second negative sample data comprises:
acquiring orders with low scores and complaints at the same time from a basic order data set as second negative sample data;
the step of sampling the first positive sample data and the third negative sample data according to a preset positive-negative sample ratio to obtain the model data includes:
limiting the proportion of the first positive sample data and the third negative sample data according to a proportion limiting formula;
the formula of the ratio limit is as follows:
wherein N ispos,Nneg,NaddRespectively representing the number of instances of the first positive sample data, the number of instances of the first negative sample data, the number of instances of the second negative sample data, neg2posRate representing the ratio of the number of instances of the first positive sample data to the number of instances of the third negative sample data, and negSampleRate representing the ratio of the first negative sample data;
the model data comprises the first positive sample data and the third negative sample data.
7. A method for identifying a swipe line, comprising:
acquiring order data to be identified;
inputting the order data to be identified into a single-file-brushing-line identification model for identification so as to obtain a single-file-brushing-line identification result;
the brush line is generated for the recognition model using the method of generating a brush line for the recognition model according to any one of claims 1 to 6.
8. A system for generating a brush line as a recognition model, comprising:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring a plurality of historical orders and constructing model data based on the historical orders;
a first construction module for constructing basic information features based on the model data;
the second construction module is used for constructing context information characteristics based on the trend of the single brushing behavior and the model data;
and the training module is used for training by adopting an ensemble learning method based on the basic information characteristics and the context information characteristics to obtain the brushing row as a recognition model.
9. The system for generating a brush line for a recognition model according to claim 8,
the second building block comprises:
the first obtaining unit is used for obtaining one from a plurality of historical orders as a current order at a time and obtaining a product and a supplier corresponding to the current order;
a second obtaining unit, configured to obtain N historical orders of the product that are adjacent to the current order and serve as first orders to be screened;
a third obtaining unit, configured to obtain N historical orders of the supplier that are adjacent to the current order and serve as a second order to be screened;
and the screening unit is used for calculating the order proportion and/or the order number meeting the preset screening condition based on the first order to be screened and the second order to be screened, and taking the order proportion and/or the order number as the context information characteristic.
10. The system for generating a swipe bank for identification models according to claim 9, wherein the contextual information features include at least one of the number of orders with the same hardware device but different user names, the proportion of orders with the same payment type, the proportion of orders with similar users, the proportion of orders with abnormal phone numbers and/or certificate numbers.
11. The system for generating a swipe line for identifying a model according to claim 10, wherein when the contextual information features include the order proportion occupied by the orders with the same payment type and/or the number of orders with the same hardware device but different user names, the filtering unit includes a corresponding calculating subunit of the order proportion occupied by the orders with the same payment type and/or a corresponding calculating subunit of the number of orders with the same hardware device but different user names; the screening unit further comprises a preprocessing subunit;
the preprocessing subunit is configured to acquire any one of the plurality of historical orders as a current order, and acquire a product and a supplier corresponding to the current order; grouping a plurality of historical orders according to the same products and the same suppliers respectively to obtain a first order group and a second order group; screening N adjacent orders which are the same as the products and suppliers of the current order from the first order group and the second order group to obtain a third order group;
the calculation subunit of the order proportion occupied by the orders with the same payment type is used for screening n orders with the same order payment mode as the current order in the third order group; calculating the order proportion occupied by the orders with the same payment type according to a context characteristic calculation formula;
the context feature calculation formula is as follows:
wherein R issamePayRepresenting the order proportion occupied by the orders with the same payment type;
the calculation subunit of the same hardware equipment but with different orders of the user name is used for screening out the order which is the same as the hardware equipment of the current order from the third order group to obtain a fifth order group; and counting the number of user names in the fifth order group as the number of orders with the same hardware equipment but different user names.
12. The system for generating a brush line for a recognition model of claim 8, wherein the first building module comprises:
the fourth acquisition unit is used for acquiring a plurality of historical orders of various channels;
the labeling unit is used for labeling each historical order to obtain first sample data, and the labeling result comprises a brushing list and a non-brushing list; the first sample data with the labeling result of the list brushing is first positive sample data, and the first sample data with the labeling result of the list brushing is first negative sample data;
the extension unit is used for carrying out negative sample extension based on the prior knowledge to obtain second negative sample data;
a merging unit, configured to merge the second negative sample data with the first negative sample data to obtain third negative sample data;
and the sampling unit is used for sampling the first positive sample data and the third negative sample data according to a preset positive-negative sample proportion to obtain the model data.
13. The system for generating a brush line for a recognition model according to claim 12,
the expansion unit is further used for acquiring orders with low scores and complaints simultaneously from a basic order data set as the second negative sample data;
the sampling unit is further configured to limit a proportion of the first positive sample data and the third negative sample data according to a formula for proportional limitation;
the formula of the ratio limit is as follows:
wherein N ispos,Nneg,NaddRespectively representing the number of instances of the first positive sample data, the number of instances of the first negative sample data, the number of instances of the second negative sample data, neg2posRate representing the ratio of the number of instances of the first positive sample data to the number of instances of the third negative sample data, and negSampleRate representing the ratio of the first negative sample data;
the model data comprises the first positive sample data and the third negative sample data.
14. A system for identifying a swipe action line, comprising:
the second acquisition module is used for acquiring order data to be identified;
the identification module is used for inputting the order data to be identified into the single-row-brushing identification model for identification so as to obtain a single-row-brushing identification result;
the swipe line is generated for the recognition model using the system for generating a swipe line for the recognition model according to any one of claims 8 to 13.
15. 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 for generating a brush line for recognition model according to any one of claims 1 to 6 or the method for recognizing a brush line for recognition according to claim 7 when executing the computer program.
16. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method of generating a brush line as claimed in any one of claims 1 to 6 as an identification model or the steps of the method as claimed in claim 7 as an identification method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911356516.9A CN111126629B (en) | 2019-12-25 | 2019-12-25 | Model generation method, brush list identification method, system, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911356516.9A CN111126629B (en) | 2019-12-25 | 2019-12-25 | Model generation method, brush list identification method, system, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111126629A true CN111126629A (en) | 2020-05-08 |
| CN111126629B CN111126629B (en) | 2023-09-19 |
Family
ID=70502074
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911356516.9A Active CN111126629B (en) | 2019-12-25 | 2019-12-25 | Model generation method, brush list identification method, system, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111126629B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112348624A (en) * | 2020-09-24 | 2021-02-09 | 北京沃东天骏信息技术有限公司 | Order processing method and device based on neural network model |
| CN113762300A (en) * | 2020-06-28 | 2021-12-07 | 北京沃东天骏信息技术有限公司 | Order classification model training method and device and order detection method and device |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017084521A1 (en) * | 2015-11-18 | 2017-05-26 | 阿里巴巴集团控股有限公司 | Order clustering method and device, and malicious information rejecting method and device |
| CN107146089A (en) * | 2017-03-29 | 2017-09-08 | 北京三快在线科技有限公司 | The single recognition methods of one kind brush and device, electronic equipment |
| CN108596632A (en) * | 2017-03-07 | 2018-09-28 | 北京嘀嘀无限科技发展有限公司 | A kind of anti-cheating recognition methods and system based on order attributes and user behavior |
| CN109345332A (en) * | 2018-08-27 | 2019-02-15 | 中国民航信息网络股份有限公司 | A kind of intelligent detecting method of Airline reservation malicious act |
| CN109509048A (en) * | 2017-09-15 | 2019-03-22 | 北京京东尚科信息技术有限公司 | The recognition methods of malice order, device, electronic equipment and storage medium |
| CN110046951A (en) * | 2018-12-29 | 2019-07-23 | 中国银联股份有限公司 | A kind of trading activity judgment method and system |
| CN110198248A (en) * | 2018-02-26 | 2019-09-03 | 北京京东尚科信息技术有限公司 | The method and apparatus for detecting IP address |
-
2019
- 2019-12-25 CN CN201911356516.9A patent/CN111126629B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017084521A1 (en) * | 2015-11-18 | 2017-05-26 | 阿里巴巴集团控股有限公司 | Order clustering method and device, and malicious information rejecting method and device |
| CN108596632A (en) * | 2017-03-07 | 2018-09-28 | 北京嘀嘀无限科技发展有限公司 | A kind of anti-cheating recognition methods and system based on order attributes and user behavior |
| CN107146089A (en) * | 2017-03-29 | 2017-09-08 | 北京三快在线科技有限公司 | The single recognition methods of one kind brush and device, electronic equipment |
| CN109509048A (en) * | 2017-09-15 | 2019-03-22 | 北京京东尚科信息技术有限公司 | The recognition methods of malice order, device, electronic equipment and storage medium |
| CN110198248A (en) * | 2018-02-26 | 2019-09-03 | 北京京东尚科信息技术有限公司 | The method and apparatus for detecting IP address |
| CN109345332A (en) * | 2018-08-27 | 2019-02-15 | 中国民航信息网络股份有限公司 | A kind of intelligent detecting method of Airline reservation malicious act |
| CN110046951A (en) * | 2018-12-29 | 2019-07-23 | 中国银联股份有限公司 | A kind of trading activity judgment method and system |
Non-Patent Citations (3)
| Title |
|---|
| LIN Y,ZHU T, WU H, ET AL: "Towards online anti-opinion spam: Spotting fake reviews from the review sequence" * |
| 刘畅: "电子商务中基于深度学习的虚假交易识别研究" * |
| 方兴林: "演化博弈视角下网络交易平台"刷单炒信"行为控制研究" * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113762300A (en) * | 2020-06-28 | 2021-12-07 | 北京沃东天骏信息技术有限公司 | Order classification model training method and device and order detection method and device |
| CN112348624A (en) * | 2020-09-24 | 2021-02-09 | 北京沃东天骏信息技术有限公司 | Order processing method and device based on neural network model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111126629B (en) | 2023-09-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20200074274A1 (en) | System and method for multi-horizon time series forecasting with dynamic temporal context learning | |
| Kotu et al. | Predictive analytics and data mining: concepts and practice with rapidminer | |
| CN107766929B (en) | Model analysis method and device | |
| CN111723292B (en) | Recommendation method, system, electronic equipment and storage medium based on graph neural network | |
| CN106909560B (en) | Interest point ordering method | |
| CN110995459B (en) | Abnormal object identification method, device, medium and electronic equipment | |
| CN110909125B (en) | Detection method of media rumor of news-level society | |
| CN112907063B (en) | Population mobility rate and house vacancy rate determining method and terminal equipment | |
| CN112925911B (en) | Complaint classification method based on multi-modal data and related equipment thereof | |
| CN111815060A (en) | Short-term load prediction method and device for power utilization area | |
| CN112036483B (en) | AutoML-based object prediction classification method, device, computer equipment and storage medium | |
| CN110858219A (en) | Logistics object information processing method and device and computer system | |
| CN111126629B (en) | Model generation method, brush list identification method, system, equipment and medium | |
| CN112988840A (en) | Time series prediction method, device, equipment and storage medium | |
| CN112614004A (en) | Method and device for processing power utilization information | |
| CN115375205A (en) | Method, device and equipment for determining water user portrait | |
| CN115545886A (en) | Overdue risk identification method, overdue risk identification device, overdue risk identification equipment and storage medium | |
| CN110334720A (en) | Feature extracting method, device, server and the storage medium of business datum | |
| CN111210332A (en) | Method and device for generating post-loan management strategy and electronic equipment | |
| CN114118570A (en) | Service data prediction method and device, electronic equipment and storage medium | |
| CN115249081A (en) | Object type prediction method and device, computer equipment and storage medium | |
| CN110717577A (en) | Time series prediction model construction method for noting regional information similarity | |
| CN116257758A (en) | Model training method, crowd expanding method, medium, device and computing equipment | |
| CN116187675A (en) | Task allocation method, device, equipment and storage medium | |
| CN112463964B (en) | Text classification and model training method, device, equipment and storage medium |
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 |