WO2019137485A1 - 一种业务分值的确定方法、装置及存储介质 - Google Patents

一种业务分值的确定方法、装置及存储介质 Download PDF

Info

Publication number
WO2019137485A1
WO2019137485A1 PCT/CN2019/071392 CN2019071392W WO2019137485A1 WO 2019137485 A1 WO2019137485 A1 WO 2019137485A1 CN 2019071392 W CN2019071392 W CN 2019071392W WO 2019137485 A1 WO2019137485 A1 WO 2019137485A1
Authority
WO
WIPO (PCT)
Prior art keywords
user
data
service
service score
score
Prior art date
Application number
PCT/CN2019/071392
Other languages
English (en)
French (fr)
Inventor
张海征
胡瑞
董雪
蒋同庆
林肯
王盛
姜凌瀚
黄超
王翔天
任玫霏
纪烨
谢育娟
鲁军宜
陈媛
Original Assignee
腾讯科技（深圳）有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 腾讯科技（深圳）有限公司 filed Critical 腾讯科技（深圳）有限公司
Publication of WO2019137485A1 publication Critical patent/WO2019137485A1/zh

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
    • 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
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G06Q30/0202—Market predictions or forecasting for commercial activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241—Advertisements
    • G06Q30/0251—Targeted advertisements

Definitions

  • the present invention relates to the field of data processing technologies, and in particular, to a technology for determining a service score.
  • Advertising is one of the most common ways of disseminating information on the Internet. More and more advertisers tend to display advertisements to users through online media platforms.
  • Advertisers can collect consumers with the same potential needs through digital marketing for a certain period of time. At present, advertisers mainly conduct customer-oriented marketing through general investment or simple crowd orientation. After the advertiser collects a large number of sales leads through the advertising collector, the customer service staff will randomly and disorderly call back the thread to achieve clue cleaning and sales follow-up.
  • the embodiment of the invention provides a method, a device and a storage medium for determining a service score.
  • advertisement placement can be performed according to a preferred group to improve the efficiency of the collection.
  • the obtained second service score has a stronger business sense and the discrimination is obvious, which can better reflect the user's purchase intention, thereby achieving priority cleaning and sales follow-up for the user with high intention to purchase, etc.
  • the practicality of the program and marketing efficiency save labor costs.
  • an embodiment of the present invention provides a method for determining a service score, including:
  • An embodiment of the present invention provides a service score determining apparatus, including:
  • An obtaining module configured to acquire target feature data of the target user
  • the acquiring module is configured to acquire, according to the target feature data and the service prediction model, a first service score corresponding to the target user, where the service prediction model is obtained by training according to a positive sample and a negative sample.
  • the positive sample is the sample data corresponding to the sold user in the sales history data
  • the negative sample is the sample data corresponding to the unsuccessful user in the sales history data;
  • a determining module configured to determine a second service score according to the service score conversion model and the first service score obtained by the obtaining module, where the second service score and the target user's purchase intention degree Positive correlation.
  • An embodiment of the present invention provides a service score determining apparatus, including: a memory, a transceiver, a processor, and a bus system;
  • the memory is used to store a program
  • the processor is configured to execute a program in the memory, including the following steps:
  • the bus system is configured to connect the memory and the processor to cause the memory and the processor to communicate.
  • An embodiment of the present invention provides a computer readable storage medium having instructions stored therein that, when run on a computer, cause the computer to perform the above method.
  • an embodiment of the present invention provides a computer program product comprising instructions that, when run on a computer, cause the computer to perform the method described above.
  • a method for determining a service score is provided. First, the target feature data of the target user is obtained, and then the first service score corresponding to the target user is obtained according to the target feature data and the service prediction model, where the service The prediction model is obtained by training according to the positive sample and the negative sample.
  • the positive sample is the sample data corresponding to the sold user in the sales history data
  • the negative sample is the sample data corresponding to the unsuccessful user in the sales history data.
  • the second service score is determined according to the business score conversion model and the first service score, wherein the second service score is positively correlated with the target user's purchase intention degree.
  • the positive and negative samples are machine-learned, the business prediction model is established, and the priority group is determined according to the first service score, so that the advertisement is delivered according to the preferred group to improve the efficiency of the collection.
  • the obtained second business score has a stronger business sense and the discrimination is obvious, which can better reflect the user's purchase intention, thereby achieving priority cleaning and sales follow-up for the user with high intention to purchase, etc., and improving the scheme. Practicality and marketing efficiency, saving labor costs.
  • FIG. 1A is a topological schematic diagram of an advertisement delivery system according to an embodiment of the present invention.
  • FIG. 1B is a schematic flowchart of advertisement delivery in an embodiment of the present invention.
  • FIG. 2 is a schematic diagram of an embodiment of an advertisement delivery policy according to an embodiment of the present invention.
  • FIG. 3 is a schematic structural diagram of an advertisement delivery system according to an embodiment of the present invention.
  • FIG. 4 is a schematic diagram of an embodiment of a method for determining a service score according to an embodiment of the present invention
  • FIG. 5 is a schematic diagram of an embodiment of a service score determining apparatus according to an embodiment of the present invention.
  • FIG. 6 is a schematic diagram of another embodiment of a service score determining apparatus according to an embodiment of the present invention.
  • FIG. 7 is a schematic structural diagram of a service score determining apparatus according to an embodiment of the present invention.
  • the embodiment of the invention provides a method, a device and a storage medium for determining a service score.
  • an advertisement can be placed according to a preferred group to improve the efficiency of the collection.
  • a priority is given to the user with high intention to purchase. Cleaning and sales follow-up, etc., improve the practicality and marketing efficiency of the program, saving labor costs.
  • FIG. 1A is a topological diagram of an advertisement delivery system according to an embodiment of the present invention. As shown in the figure, a server and a plurality of terminals are included in an advertisement delivery system, and the server first collects a large amount of sales history data, such as a transaction.
  • the server can use the sales history data as sample data, and train the sample data to obtain a business prediction model, and predict which users are high-latency users through the business prediction model.
  • the server pushes advertisements to terminals used by these high-latency users, such as terminal A, terminal B, terminal C, terminal D, and terminal E.
  • FIG. 1B is a schematic flowchart of advertisement delivery according to an embodiment of the present invention.
  • the steps S1, S2, and S3 are included, and the advertisement exposure in step S1 is also performed. That is, the advertisement is placed on each platform, and the advertisement click in step S2 is the behavior of the user clicking and watching the advertisement on the platform.
  • the user behavior can be converted into a sales thread, wherein the sales thread can be a marketing parameter for the user to fill in the personal data in the big data era to enable the advertiser to collect the customer.
  • step S4 the personnel of the telemarketing center perform the clue cleaning, and the clue cleaning is to return the visitor by the advertiser's customer service personnel through the contact information filled out by the user, inquiring the user to purchase the intention, excluding the clue without purchase intention, and having the purchase intention
  • step S5 the terminal salesperson performs follow-up, that is, the salesperson of the advertiser performs a one-to-one sales service to the user having the purchase intention.
  • FIG. 2 is a schematic diagram of an embodiment of an advertisement delivery strategy according to an embodiment of the present invention.
  • the initial model training needs to be cold-started, that is, access-related.
  • the sample data is preprocessed, and the sample data is divided into a positive sample and a negative sample, and the business prediction model is trained by a machine learning method.
  • the high-potential crowd is preferred, that is, based on the business forecasting model, the score obtained by the scoring is the first business score, which is also the main link of the entire advertising strategy. .
  • the high-potential clue is selected, and the score obtained based on the business prediction model is scored by the business score conversion model to obtain the business score.
  • the business score obtained by the scoring step is The second business score.
  • the operation guidance and sales output business scores can help the sales center and terminal sales improve marketing efficiency.
  • the fifth step of the advertising strategy the real transaction situation is fed back, and the business forecasting model is optimized based on the real transaction data feedback.
  • FIG. 3 is a schematic structural diagram of an advertisement delivery system according to an embodiment of the present invention.
  • five modules are included in an advertisement delivery system, and specifically, a service prediction model training is performed in a first module.
  • the raw data obtained is preprocessed, and then the processed data is divided into a positive sample and a negative sample.
  • Machine learning is performed on positive and negative samples, and a business prediction model and a business score conversion model are obtained.
  • the second module the data of the full amount of users and the customer relationship management (CRM) data are input into the business prediction model, and the data of the full amount of users and the data of the advertiser CRM may be referred to as feature data, and the business prediction is performed.
  • CRM customer relationship management
  • the model outputs the first business score of each user to determine the ranking of the delivery group, and then the corresponding advertisements are delivered to the people, and the sales leads can be continuously obtained after the advertisement is placed.
  • the first business is divided. The value is input to the business score conversion model, and then the second business score is obtained. According to the second business score, the user purchase intention can be sorted and graded, and the sales lead is added to the business score conversion model and the business prediction model. Used to further train the business score conversion model and the business prediction model.
  • operational guidance and sales are based on user ranking and grading, and finally the real transaction data is fed back in the fifth module.
  • the advertiser obtains a sales lead in the promotion of the advertiser, which can be the user's basic data (such as age, gender, occupation, etc.) filled in by the user, and at the same time, through the Data-Management Platform (DMP).
  • the real transaction data is pushed to the prediction model module in real time through the application programming interface (API).
  • API application programming interface
  • the model predicts the purchase intention based on the clue identifier and the FTRL framework, and the business score result is fed back to the advertiser in real time through the API.
  • the advertiser passes back the personal data and purchase behavior data of the DMP to the model training module to form a closed loop of data, and conducts a purchase intention prediction model (ie, a business prediction model). )optimization.
  • an embodiment of the method for determining the service score in the embodiment of the present invention includes:
  • the service score determining apparatus first acquires target feature data of the target user, wherein the target feature data includes at least one of basic data, semantic data, and behavior data of the target user, and the basic data of the target user includes but not only It is limited to at least one of the target user's age, gender, region, education, occupation, and marriage status.
  • Semantic data includes keyword information that the user enters or reads.
  • Behavioral data includes, but is not limited to, the installation of the application, the active index of the application, and the operational behavior of the QQ space.
  • the service score determining device inputs the target feature data of the target user to the service prediction model, and outputs the first service score corresponding to the target user through the service prediction model.
  • the business prediction model here is obtained according to the training of the positive sample and the negative sample.
  • the positive sample is the sample data corresponding to the sold users in the sales history data
  • the negative sample is the sample data corresponding to the unsuccessful users in the sales history data.
  • the positive sample and the negative sample may respectively comprise a plurality of different data.
  • the positive sample may include first user basic data, first semantic data, first behavior data; and the negative sample may include a second user basic Data, second semantic data, and second behavior data.
  • the service score determining apparatus further converts the first service score into the second service score according to the service score conversion model, and the service score conversion model may be a linear transformation model.
  • the second business score is positively related to the purchase intention of the target user. That is to say, the higher the second business score means that the target user may purchase the commodity more likely, and the second business score is better.
  • the interpretability for example, 50 points and 60 points have different meanings.
  • the second service score can flexibly adjust the score range and density according to the specific business application scenario, and facilitate the classification of the target user according to the second service score.
  • a method for determining a service score is provided. First, the target feature data of the target user is obtained, and then the first service score corresponding to the target user is obtained according to the target feature data and the service prediction model, where the service The prediction model is obtained by training according to the positive sample and the negative sample.
  • the positive sample is the sample data corresponding to the sold user in the sales history data
  • the negative sample is the sample data corresponding to the unsuccessful user in the sales history data.
  • the second service score is determined according to the business value conversion model and the first business score, wherein the second business score is positively correlated with the target user's purchase intention degree.
  • the first service segment corresponding to the target feature data is obtained by using the service prediction model.
  • the service prediction model Before the value, it can also include:
  • the purchase result corresponding to the positive sample, the positive sample, the negative sample, and the negative sample are trained to obtain a business prediction model.
  • the service score determining apparatus needs to obtain the service prediction model before acquiring the first service score.
  • a specific training method is training through Logistic Regression (LR).
  • LR is a classification model in machine learning. Due to its simplicity and efficiency, it is widely used in practice.
  • the LR belongs to supervised learning. Therefore, before using these algorithms, it is necessary to collect a batch of labeled sample data as a training set.
  • the sample data includes a positive sample and a negative sample, specifically the first user basic data, first Semantic data, first behavior data, second user basic data, second semantic data, and second behavior data.
  • the first user basic data is the personal information of the sold user
  • the first semantic data is the keyword information of the sold user
  • the first behavior data is the operation information of the sold user.
  • the second user basic data is the personal information of the unpaid user
  • the second semantic data is the keyword information of the unpaid user
  • the second behavior data is the operation information of the unpaid user.
  • Some of the labeled sample data can be obtained from the database (such as user clicks or purchases), some of the labeled sample data can be obtained from the information filled in by the user (such as gender), and some sample data is marked by some Manually labeled.
  • LR is often used to solve dichotomy problems, such as predicting whether a user clicks on a particular ad, or predicting whether a user is buying a particular branded car, or predicting whether a user is interested in learning IELTS. Therefore, our sample data can be expressed as:
  • x i represents at least one of the first user basic data, the first semantic data, and the first behavior data (or at least one of the second user basic data, the second semantic data, and the second behavior data) M-dimensional vector
  • y indicates the purchase result. When y is equal to 0, it indicates that it has not been purchased. When y is equal to 1, it indicates that it has been purchased. Please refer to Table 1.
  • Table 1 is an indication of sample data.
  • the sample data of the purchased car represents a positive sample
  • the sample data of the unpurchased car represents a negative sample
  • x i in the first row of Table 1 predicting the possibility of the user purchasing a car
  • the process of training the business prediction model is actually the process of solving the likelihood ratio parameter ⁇ in the LR by using the maximum likelihood ratio estimation value, specifically:
  • Gradient Descent can be used to optimize the parameters to minimize the loss, so that the optimal likelihood ratio parameter ⁇ can be obtained.
  • Other convex optimization methods have a conjugate gradient drop, Newton method.
  • a method of training to obtain a business prediction model is introduced, that is, a logistic regression training is performed on a positive sample and a negative sample, and a business prediction model can be obtained.
  • the accuracy of the training can be improved, and at the same time, the logistic regression training is applicable to the independent and categorical independent variables, and is easy to use and explain.
  • the positive sample may include the first user basic data, The first semantic data and the first behavior data, wherein the first user basic data is the personal information of the sold user, the first semantic data is the keyword information of the sold user, and the first behavior data is the operation information of the sold user;
  • the negative sample may include second user basic data, second semantic data, and second behavior data, wherein the second user basic data is personal information of the unpaid user, the second semantic data is keyword information of the unpaid user, and the second The behavior data is the operation information of the unpaid user.
  • the sample data trained by the service score determining device may be derived from a first user feature database in the server and/or a second user feature database provided by the user.
  • the first user feature library may be a user feature library provided by Tencent, and the feature database is characterized by large and comprehensive data.
  • the second user feature database may be the own feature data provided by the advertiser, such as the intent city and the user history purchase record, etc., and the data in the first user feature database and the data in the second user feature database are spliced together. Positive and negative samples are trained.
  • the splicing process here may be to summarize data with the same user identifier, and the user identifier may be a mobile phone number or a QQ number, etc., and more data is mapped and summarized, which can increase the training dimension of the data amount, thereby improving the sample. Corrected space.
  • the user basic data includes a basic attribute class, a socioeconomic status class, a location based service (LBS) class, a device class, a business interest class, and a vertical industry tag class
  • the semantic data includes a semantic class and behavior data. Includes business behavior classes.
  • the basic attribute class includes, but is not limited to, age, gender, geography, education, occupation, and status of marriage; socioeconomic status categories include, but are not limited to, resident countries, resident provinces, and resident cities; Not limited to operating system, operator, equipment type fee, purchase situation, car purchase situation, traveler, consumption record and monthly service; LBS category includes but not limited to online scene; commercial interest category includes but is not limited to browsing behavior; vertical
  • the industry label category can be "car _ car sneak _ specific to the model", or "car _ car crowd _ specific to the model", or "car _ used car crowd _ specific to the model” and so on.
  • the semantic class includes keyword information input or read by the user. For example, the user inputs “how is the effect of red wine papaya soup", firstly, the word segmentation process is performed, and “red wine”, “papaya”, “soup”, “effect” are obtained. And “how”, the keywords are “red wine papaya soup”, “red wine papaya”, “papaya soup”, “red wine” and “papaya”, according to the keyword mapping to the corresponding topic can be “beauty slimming", “beauty Plastic surgery, food and beverage, and food.
  • Behavioral data includes, but is not limited to, the installation of the application, the active index of the application, and the operational behavior of the QQ space.
  • the third optional embodiment of determining the service score provided by the embodiment of the present invention before obtaining the positive sample and the negative sample, include:
  • the preprocessed negative samples are preprocessed to obtain negative samples, wherein the preprocessing includes at least one of deduplication processing, active sample extraction, and associated sample combination.
  • the service score determining apparatus first obtains the pre-processed positive sample and the pre-processed negative sample, and then preprocesses the pre-processed positive sample and the pre-processed negative sample, respectively.
  • the manner of pre-processing includes, but is not limited to, at least one of de-duplication processing, active sample extraction, and associated sample merging.
  • the first type of de-duplication is to remove the duplicate data. Because in many cases, multiple identical data corresponding to the same user may be obtained, but training multiple repeated data will reduce the accuracy of the model, so it is necessary to eliminate these repeated data.
  • the second active sample extraction that is, the data corresponding to the user with higher activity is taken as the sample data. This is because the sample data should be highly representative, and the high user activity indicates that the user has a larger amount of data and has greater reliability.
  • the third associated sample is merged.
  • One user may be bound to multiple accounts at the same time.
  • user A has a QQ number and a micro signal, but both accounts are bound to the mobile phone number of user A, and then the mobile phone can be determined.
  • the QQ number and the micro signal bound to the number belong to User A, so the data on the QQ can be combined with the data on the WeChat to obtain a sample data.
  • a plurality of accounts have a binding relationship with a user identifier.
  • the user identifier herein may be a mobile phone number (International Mobile Equipment Identity, IMEI). ), QQ number or Media Access Control (MAC) address.
  • IMEI International Mobile Equipment Identity
  • MAC Media Access Control
  • IMEI is also known as the international mobile device identification, which is the unique identification number of the mobile phone.
  • a MAC address also known as a physical address or hardware address, can be used to define the location of a network device and is globally unique.
  • the positive sample and the negative sample need to be preprocessed, for example, at least one of deduplication processing, active sample extraction, and associated sample merging.
  • deduplication processing active sample extraction
  • associated sample merging the sample data closer to the actual situation can be obtained, thereby improving the accuracy of the model training.
  • the first corresponding to the target feature data is obtained by using the service prediction model.
  • Business scores which can include:
  • the first business score is calculated using the business prediction model:
  • P(y 1
  • x; 0) represents the first service score
  • the ⁇ represents a likelihood ratio parameter
  • the X represents the target feature data
  • the y represents a probability of a purchase result Value
  • the T represents a transposed matrix
  • the K-fold cross-validation method is used to perform LR training on the sample feature data to obtain an optimal model, that is, a business prediction model is obtained.
  • LR is a generalized linear regression analysis model, which is commonly used to solve classification problems.
  • the FTRL training framework is an algorithm for batch processing of very large data sets and online data streams.
  • the decision function obtained by LR training is
  • X represents target feature data
  • the target feature data includes, but is not limited to, user basic data, semantic data, and behavior data of the target user.
  • P(y 1
  • x; 0) represents the purchase intention probability of the target feature data at the first service score, that is, the possibility that the target feature data X purchases the commodity under the parameter ⁇ .
  • the K-fold cross-validation method can be used for solving ⁇ .
  • a method for calculating a first service score is provided, that is, a required first service score can be calculated by using a service prediction model. In the above manner, the practicability and feasibility of the solution can be improved.
  • the method for determining the service score provided by the embodiment of the present invention is determined according to the service score conversion model and the first service score.
  • the second business score can include:
  • the score indicates the second service score
  • the average indicates that the first service score is equal to the expected score when pos_neg_ratio
  • the density is used to indicate the density of the second service score.
  • the probability represents the first service score
  • the pos_neg_ratio represents a ratio between the positive sample and the negative sample.
  • the first service score is converted into the second service score by using the service score conversion model. Specifically, the original first service score outputted by the service prediction model is converted to obtain a second service score with significant business meaning and distinctiveness. ;
  • the average can be flexibly adjusted according to the specific business background.
  • the average can be set to the average of the expected scores, for example, the expected score range is [0, 100]. In general, the average can be set to 50. It should be noted that the expected range of values is a range that can be preset. In addition to [0,100], it can be [0,150], or [10,160], etc. The invention is not to be construed as limiting the invention.
  • Density is used to control the density of the second service score.
  • the density indicates the score when the first service score is doubled in the service scenario, that is, the resolution of the final score. For example, if the density is 10, the second is described.
  • Each additional 10 points for the business score indicates a doubling of the probability of a purchase rate. Density can be achieved by trying different values, observing which value can get better discrimination and business meaning, and the density generally does not exceed the range of expected scores. For example, the expected score range is [0,100], you can try 5, 10, 15, 20 and other density for observation tuning.
  • Pos_neg_ratio represents the ratio between the number of positive samples and the number of negative samples in the training sample, such as a total of 10 samples, of which 6 samples are purchased (ie, positive samples), and 4 samples are unpurchased vehicles. (ie negative sample), then pos_neg_ratio is 1.5.
  • the ratio pos_neg_ratio between the positive and negative samples is 2.3%.
  • the second service score is 60 points, it means that the user's purchase rate is 9.2. %.
  • the first hypothesis is to set a specific positive and negative sample ratio pos_neg_ratio to a specific expected score average;
  • the second hypothesis is to determine the fractionality of the ratio doubling.
  • density is equal to 10
  • Average+density A+B+log(2*pos_neg_ratio);
  • the expected score range can be 0 to 100, while the average can take 50, and the density can take the value 10.
  • a method for calculating a second service score is provided, that is, a required second service score can be calculated by using a service score conversion model. In the above manner, the practicability and feasibility of the solution can be improved.
  • the target user may be any user that needs to perform service analysis, that is, for each target user.
  • the steps of the foregoing 101-103 may be performed, so that the service score determining apparatus may obtain the second service score of each target user, and then sort the different target users according to the size of the second service score, and the sort result may reflect The order of the purchase intentions of different target users.
  • the second service score corresponding to each user is calculated, and if the second service score of 20 users is reached, Threshold, then these 20 users are high-latency users and can be used as high-potential clues.
  • the second service scores of the 20 users are sorted, and the sorting result can reflect the order of purchase intention degree of different target users, and the customer service personnel can perform the thread cleaning according to the order of the second service scores from high to low. That is, the target user is subjected to clue cleaning and sales follow-up according to the order of purchase intention from large to small.
  • the users after obtaining the service scores of the multiple users, the users may be sorted in order from the highest to the lowest of the second service scores.
  • the customer service personnel can perform the thread cleaning according to the second service score, thereby achieving efficient follow-up of the high purchase intention user, thereby improving sales. The rate of achievement.
  • FIG. 5 is a schematic diagram of an embodiment of a service score determining apparatus according to an embodiment of the present invention.
  • the service score determining apparatus 20 includes:
  • the obtaining module 201 is configured to acquire target feature data of the target user.
  • the obtaining module 201 is configured to acquire, according to the target feature data and the service prediction model, a first service score corresponding to the target user, where the service prediction model is obtained by training according to a positive sample and a negative sample,
  • the positive sample is sample data corresponding to the sold user in the sales history data
  • the negative sample is sample data corresponding to the unsuccessful user in the sales history data
  • a determining module 202 configured to determine a second service score according to the service score conversion model and the first service score obtained by the obtaining module 201, where the second service score and the target user purchase Intentionality is positively correlated.
  • the acquiring module 201 acquires the target feature data of the target user, and the acquiring module 201 inputs the target feature data into the service prediction model, thereby outputting the first service score corresponding to the target user, and determining the module.
  • the first service score obtained by the obtaining module 201 is converted into a second service score by the service score conversion model.
  • a service score determining apparatus which first acquires target feature data of a target user, and then acquires a first service score corresponding to the target user according to the target feature data and the service prediction model, where the service prediction The model is obtained according to the training of the positive sample and the negative sample.
  • the positive sample is the sample data corresponding to the sold user in the sales history data
  • the negative sample is the sample data corresponding to the unsuccessful user in the sales history data.
  • the second service score is determined according to the business score conversion model and the first service score, wherein the second service score is positively correlated with the target user's purchase intention degree.
  • the service score determining apparatus 20 further Include a training module 203;
  • the obtaining module 201 is further configured to acquire the positive sample and the negative sample before acquiring the first service score corresponding to the target user according to the target feature data and the service prediction model;
  • the training module 203 is configured to train the positive sample acquired by the obtaining module 201, the purchase result corresponding to the positive sample, the negative sample, and the purchase result corresponding to the negative sample to obtain The business prediction model.
  • the manner of training the business prediction model is introduced, that is, the logistic regression training is performed on the positive sample and the negative sample, and the business prediction model can be obtained.
  • the accuracy of the training can be improved, and at the same time, the logistic regression training is applicable to the independent and categorical independent variables, and is easy to use and explain.
  • the positive sample includes the first user basic data, the first semantic data, and the first behavior data, wherein the first user basic data is personal information of an already-paid user, and the first semantic data For the keyword information of the sold user, the first behavior data is operation information of the sold user;
  • the negative sample includes the second user basic data, the second semantic data, and the second behavior data, wherein the second user basic data is personal information of an unpaid user, and the second semantic data For the keyword information of the unpaid user, the second behavior data is operation information of the unpaid user.
  • the obtaining module 201 is specifically configured to calculate the first service score by using the service prediction model:
  • P(y 1
  • x; 0) represents the first service score
  • the ⁇ represents a likelihood ratio parameter
  • the X represents the target feature data
  • the y represents a probability of a purchase result Value
  • the T represents a transposed matrix
  • a method for calculating the first service score is provided, that is, the required first service score can be calculated by using the service prediction model. In the above manner, the practicability and feasibility of the solution can be improved.
  • the determining module 202 calculates the second service score by using the service score conversion model:
  • the score indicates the second service score
  • the average indicates that the first service score is equal to the expected score when pos_neg_ratio
  • the density is used to indicate the density of the second service score.
  • the probability represents the first service score
  • the pos_neg_ratio represents a ratio between the positive sample and the negative sample.
  • a method for calculating a second service score is provided, that is, a required second service score can be calculated by using a service score conversion model. In the above manner, the practicability and feasibility of the solution can be improved.
  • the users after obtaining the second service scores of the multiple users, the users may be sorted according to the order of the second service scores from high to low.
  • the customer service personnel can perform the thread cleaning according to the second service score, thereby achieving efficient follow-up of the high purchase intention user, thereby improving sales. The rate of achievement.
  • the embodiment of the present invention further provides another service score determining apparatus.
  • FIG. 7 for the convenience of description, only parts related to the embodiment of the present invention are shown. Without specific details, please refer to the present invention.
  • the terminal may be any terminal device including a mobile phone, a tablet computer, a personal digital assistant (PDA), a point of sales (POS), a car computer, and the like, and the terminal is a mobile phone as an example:
  • FIG. 7 is a block diagram showing a partial structure of a mobile phone related to a terminal provided by an embodiment of the present invention.
  • the mobile phone includes: a radio frequency (RF) circuit 310, a memory 320, an input unit 330, a display unit 340, a sensor 350, an audio circuit 360, a wireless fidelity (WiFi) module 370, and a processor 380. And power supply 390 and other components.
  • RF radio frequency
  • the RF circuit 310 can be used for transmitting and receiving information or during a call, and receiving and transmitting the signal. Specifically, after receiving the downlink information of the base station, the processor 380 processes the data. In addition, the uplink data is designed to be sent to the base station.
  • RF circuit 310 includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a Low Noise Amplifier (LNA), a duplexer, and the like.
  • LNA Low Noise Amplifier
  • RF circuitry 310 can also communicate with the network and other devices via wireless communication. The above wireless communication may use any communication standard or protocol, including but not limited to Global System of Mobile communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (Code Division). Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), E-mail, Short Messaging Service (SMS), and the like.
  • GSM Global System of Mobile communication
  • the memory 320 can be used to store software programs and modules, and the processor 380 executes various functional applications and data processing of the mobile phone by running software programs and modules stored in the memory 320.
  • the memory 320 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application required for at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may be stored according to Data created by the use of the mobile phone (such as audio data, phone book, etc.).
  • memory 320 can include high speed random access memory, and can also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device.
  • the input unit 330 can be configured to receive input numeric or character information and to generate key signal inputs related to user settings and function controls of the handset.
  • the input unit 330 may include a touch panel 331 and other input devices 332.
  • the touch panel 331 also referred to as a touch screen, can collect touch operations on or near the user (such as a user using a finger, a stylus, or the like on the touch panel 331 or near the touch panel 331 Operation), and drive the corresponding connecting device according to a preset program.
  • the touch panel 331 can include two parts: a touch detection device and a touch controller.
  • the touch detection device detects the touch orientation of the user, and detects a signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts the touch information into contact coordinates, and sends the touch information.
  • the processor 380 is provided and can receive commands from the processor 380 and execute them.
  • the touch panel 331 can be implemented in various types such as resistive, capacitive, infrared, and surface acoustic waves.
  • the input unit 330 may also include other input devices 332.
  • other input devices 332 may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control buttons, switch buttons, etc.), trackballs, mice, joysticks, and the like.
  • the display unit 340 can be used to display information input by the user or information provided to the user as well as various menus of the mobile phone.
  • the display unit 340 can include a display panel 341.
  • the display panel 341 can be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), or the like.
  • the touch panel 331 can cover the display panel 341. When the touch panel 331 detects a touch operation on or near it, the touch panel 331 transmits to the processor 380 to determine the type of the touch event, and then the processor 380 according to the touch event. The type provides a corresponding visual output on display panel 341.
  • the touch panel 331 and the display panel 341 are used as two independent components to implement the input and input functions of the mobile phone in FIG. 7, in some embodiments, the touch panel 331 can be integrated with the display panel 341. Realize the input and output functions of the phone.
  • the handset can also include at least one type of sensor 350, such as a light sensor, motion sensor, and other sensors.
  • the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 341 according to the brightness of the ambient light, and the proximity sensor may close the display panel 341 and/or when the mobile phone moves to the ear. Or backlight.
  • the accelerometer sensor can detect the magnitude of acceleration in all directions (usually three axes). When it is stationary, it can detect the magnitude and direction of gravity.
  • the mobile phone can be used to identify the gesture of the mobile phone (such as horizontal and vertical screen switching, related Game, magnetometer attitude calibration), vibration recognition related functions (such as pedometer, tapping), etc.; as for the mobile phone can also be configured with gyroscopes, barometers, hygrometers, thermometers, infrared sensors and other sensors, no longer Narration.
  • the gesture of the mobile phone such as horizontal and vertical screen switching, related Game, magnetometer attitude calibration
  • vibration recognition related functions such as pedometer, tapping
  • the mobile phone can also be configured with gyroscopes, barometers, hygrometers, thermometers, infrared sensors and other sensors, no longer Narration.
  • the audio circuit 360, the speaker 361, and the microphone 362 provide an audio interface between the user and the handset.
  • the audio circuit 360 can transmit the converted electrical data of the received audio data to the speaker 361 for conversion to the sound signal output by the speaker 361; on the other hand, the microphone 362 converts the collected sound signal into an electrical signal, by the audio circuit 360. After receiving, it is converted into audio data, and then processed by the audio data output processor 380, sent to the other mobile phone via the RF circuit 310, or outputted to the memory 320 for further processing.
  • WiFi is a short-range wireless transmission technology.
  • the mobile phone can help users to send and receive emails, browse web pages and access streaming media through the WiFi module 370, which provides users with wireless broadband Internet access.
  • FIG. 7 shows the WiFi module 370, it can be understood that it does not belong to the essential configuration of the mobile phone, and can be omitted as needed within the scope of not changing the essence of the invention.
  • the processor 380 is the control center of the handset, which connects various portions of the entire handset using various interfaces and lines, by executing or executing software programs and/or modules stored in the memory 320, and invoking data stored in the memory 320, The phone's various functions and processing data, so that the overall monitoring of the phone.
  • the processor 380 may include one or more processing units; optionally, the processor 380 may integrate an application processor and a modem processor, where the application processor mainly processes an operating system, a user interface, and an application. Etc.
  • the modem processor primarily handles wireless communications. It will be appreciated that the above described modem processor may also not be integrated into the processor 380.
  • the handset also includes a power source 390 (such as a battery) that powers the various components.
  • a power source can be logically coupled to the processor 380 through a power management system to manage charging, discharging, and power management functions through the power management system.
  • the mobile phone may further include a camera, a Bluetooth module, and the like, and details are not described herein again.
  • the processor 380 included in the terminal further has the following functions:
  • processor 380 is further configured to perform the following steps:
  • processor 380 is further configured to perform the following steps:
  • the negative sample to be processed is preprocessed to obtain the negative sample, wherein the preprocessing includes at least one of deduplication processing, active sample extraction, and associated sample combination.
  • processor 380 is specifically configured to perform the following steps:
  • P(y 1
  • x; 0) represents the first service score
  • the ⁇ represents a likelihood ratio parameter
  • the X represents the target feature data
  • the y represents a probability of a purchase result Value
  • the T represents a transposed matrix
  • processor 380 is specifically configured to perform the following steps:
  • the score indicates the second service score
  • the average indicates that the first service score is equal to the expected score when pos_neg_ratio
  • the density is used to indicate the density of the second service score.
  • the probability represents the first service score
  • the pos_neg_ratio represents a ratio between the positive sample and the negative sample.
  • the disclosed system, apparatus, and method may be implemented in other manners.
  • the device embodiments described above are merely illustrative.
  • the division of the unit is only a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.
  • the integrated unit if implemented in the form of a software functional unit and sold or used as a standalone product, may be stored in a computer readable storage medium.
  • the technical solution of the present invention which is essential or contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product stored in a storage medium.
  • a number of instructions are included to cause a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention.
  • the foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like, which can store program code. .

Abstract

本发明实施例公开了一种业务分值的确定方法、装置及存储介质，该方法获取第一目标用户的目标特征数据；通过根据目标特征数据和业务预测模型获取目标用户所对应的第一业务分值，根据第一业务分值可以确定出优先人群；根据业务分值转换模型以及所述第一业务分值确定第二业务分值，其中，所述第二业务分值与目标用户的购买意向度呈正相关。可见，本发明可以根据优选人群进行广告投放，以提升集客效率。另外，得到的第二业务分值具有更强的业务意义且区分度明显，更能体现用户购买意向，从而实现优先向购买意向度高的用户进行线索清洗和销售跟进等，提升了方案的实用性和营销效率，节省人力成本。

Description

一种业务分值的确定方法、装置及存储介质
本申请要求于2018年1月12日提交中国专利局、申请号201810033512.6、申请名称为“一种业务分值的确定方法、业务分值确定装置及介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本发明涉及数据处理技术领域，尤其涉及业务分值的确定技术。
背景技术
广告投放是互联网最常见的信息传播方式之一，越来越多的广告主趋向于通过网络媒体平台向用户展示广告。
广告主可以在一定时期内通过数字营销手段，将拥有相同潜在需求的消费者进行收集。目前，广告主主要通过通投或者简单人群定向来进行集客式营销。在广告主通过广告集客收集到大量销售线索之后，其客服人员会随机无序地对线索进行电话回访，以实现线索清洗和销售跟进。
然而，通过通投或者简单人群定向来进行集客式营销会面临集客效率低下的问题。很多时候，大量的广告投放并不能带来大量的高质量销售线索，即使很多人留下了个人信息成为线索，但其实他们并没有真正购买商品的意向。与此同时，随机人工电话回访的线索清洗方式，需要大量的人力成本，从而降低了方案的实用性。
发明内容
本发明实施例提供了一种业务分值的确定方法、装置及存储介质，一方面可以根据优选人群进行广告投放，以提升集客效率。另一方面，得到的第二业务分值具有更强的业务意义且区分度明显，更能体现用户购买意向，从而实现优先向购买意向度高的用户进行线索清洗和销售跟进等，提升了方案的实用性和营销效率，节省人力成本。
有鉴于此，本发明实施例一方面提供一种业务分值的确定方法，包括：
获取目标用户的目标特征数据；
根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值，其中，所述业务预测模型为根据正样本和负样本训练得到的，所述正样本为销售历史数据中已成交用户对应的样本数据，所述负样本为销售历史数据中未成交用户对应的样本数据；
根据业务分值转换模型以及所述第一业务分值确定第二业务分值，其中，所述第二业务分值与所述目标用户的购买意向度呈正相关。
本发明实施例一方面提供了一种业务分值确定装置，包括：
获取模块，用于获取目标用户的目标特征数据；
所述获取模块，用于根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值，其中，所述业务预测模型为根据正样本和负样本训练得到的，所述正样本为销售历史数据中已成交用户对应的样本数据，所述负样本为销售历史数据中未成交用户对应的样本数据；
确定模块，用于根据业务分值转换模型以及所述获取模块获取的所述第一业务分值确定第二业务分值，其中，所述第二业务分值与所述目标用户的购买意向度呈正相关。
本发明实施例一方面提供了一种业务分值确定装置，包括：存储器、收发器、处理器以及总线系统；
其中，所述存储器用于存储程序；
所述处理器用于执行所述存储器中的程序，包括如下步骤：
获取目标用户的目标特征数据；
根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值，其中，所述业务预测模型为根据正样本和负样本训练得到的，所述正样本为销售历史数据中已成交用户对应的样本数据，所述负样本为销售历史数据中未成交用户对应的样本数据；
根据业务分值转换模型以及所述第一业务分值确定第二业务分值，其中，所述第二业务分值与所述目标用户的购买意向度呈正相关；
所述总线系统用于连接所述存储器以及所述处理器，以使所述存储器以及所述处理器进行通信。
本发明实施例一方面提供了一种计算机可读存储介质，所述计算机可读存储介质中存储有指令，当其在计算机上运行时，使得计算机执行上述的方法。
本发明实施例一方面提供了一种计算机程序产品，包括指令，当其在计算机上运行时，使得计算机执行上述的方法。
从以上技术方案可以看出，本发明实施例具有以下优点：
本发明实施例中，提供了一种业务分值的确定方法，首先获取目标用户的目标特征数据，然后根据目标特征数据和业务预测模型获取目标用户所对应的第一业务分值，其中，业务预测模型为根据正样本和负样本训练得到的，正样本为销售历史数据中已成交用户对应的样本数据，负样本为销售历史数据中未成交用户对应的样本数据。接下来根据业务分值转换模型以及第一业务分值确定第二业务分值，其中，第二业务分值与目标用户的购买意向度呈正相关。通过上述方式，对正样本和负样本进行机器学习，建立业务预测模型， 进而根据第一业务分值确定出优先人群，从而根据优选人群进行广告投放，以提升集客效率。另外，得到的第二业务分值具有更强的业务意义且区分度明显，更能体现用户购买意向，从而实现优先向购买意向度高的用户进行线索清洗和销售跟进等，提升了方案的实用性和营销效率，节省人力成本。
附图说明
图1A为本发明实施例中广告投放系统的一个拓扑示意图；
图1B为本发明实施例中广告投放的一个流程示意图；
图2为本发明实施例中广告投放策略的一个实施例示意图；
图3为本发明实施例中广告投放系统的一个架构示意图；
图4为本发明实施例中业务分值的确定方法一个实施例示意图；
图5为本发明实施例中业务分值确定装置一个实施例示意图；
图6为本发明实施例中业务分值确定装置另一个实施例示意图；
图7为本发明实施例中业务分值确定装置一个结构示意图。
具体实施方式
本发明实施例提供了一种业务分值的确定方法、装置及存储介质，一方面可以根据优选人群进行广告投放，以提升集客效率，另一方面，实现优先向购买意向度高的用户进行线索清洗和销售跟进等，提升了方案的实用性和营销效率，节省人力成本。
本发明的说明书和权利要求书及上述附图中的术语“第一”、“第二”、“第三”、“第四”等(如果存在)是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本发明的实施例例如能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
应理解，本发明实施例可应用于广告投放的场景，主要用于获得更为贴近实际情况的购买意向，并且根据推广需求产出相应意向度范围内的人群，并对这些人群进行实际广告的投放，以此来提升广告推广的集客效率。请参阅图1A，图1A为本发明实施例中广告投放系统的一个拓扑示意图，如图所示，在广告投放系统中包括服务器以及多个终端，服务器首先收集大量的销售历史数据，比如已成交用户的个人信息，已成交用户的关键字信息，已成交用户的操作信息，未成交用户的个人信息，未成交用户的关键字信息和未成交用户的操作信息。服务器可以将这些销售历史数据作为样本数据，对这些样本数据进行训练后 得到业务预测模型，通过该业务预测模型预测哪些用户是高潜用户。服务器向这些高潜用户所使用的终端(如终端A、终端B、终端C、终端D和终端E)推送广告。
为了便于介绍，请参阅图1B，图1B为本发明实施例中广告投放的一个流程示意图，如图所示，在广告投放阶段包括步骤S1、步骤S2和步骤S3，步骤S1中的广告曝光也就是在各大平台上投放广告，步骤S2中的广告点击即为用户在平台上点击观看广告的行为。步骤S3中可以将用户行为转化为销售线索，其中，销售线索可以是一种在大数据时代用户通过在线填写个人资料使广告主集客的营销参数。在步骤S4中，电话销售中心的人员进行线索清洗，线索清洗即为广告主客服人员通过用户填写的联系方式进行回访，询问用户购买意向，将无购买意向的线索进行排除，并将有购买意向线索下发销售人员的过程。步骤S5中，终端销售人员进行跟进，也就是广告主的销售人员对具有购买意向用户进行一对一的销售服务。
请参阅图2，图2为本发明实施例中广告投放策略的一个实施例示意图，如图所示，在广告投放策略的第一步中，需要先冷启动初始模型训练，即接入相关的样本数据并进行预处理，将这些样本数据分成正样本和负样本，通过机器学习的方法训练业务预测模型。在广告投放策略的第二步中，优选高潜投放的人群，也就是基于业务预测模型进行打分，这一步打分所得到的分数即为第一业务分值，这也是整个广告投放策略的主要环节。在广告投放策略的第三步中，优选高潜线索，基于业务预测模型所得到的分数，再通过业务分值转换模型进行打分，得到业务分值，这一步打分所得到的业务分值即为第二业务分值。在广告投放策略的第四步中，根据高潜线索和高潜投放的人群进行运营引导及销售，输出业务分值可以帮助电销中心和终端销售提升营销效率。在广告投放策略的第五步中，反馈真实的成交情况，并基于真实成交数据反馈优化业务预测模型。
请参阅图3，图3为本发明实施例中广告投放系统的一个架构示意图，如图所示，在广告投放系统中包括五大模块，具体地，在第一个模块中执行业务预测模型的训练，将获取到的原始数据先进行预处理，然后将处理后的数据划分为正样本和负样本。对正样本和负样本进行机器学习，得到业务预测模型与业务分值转换模型。在第二个模块中，将全量用户的数据和广告主客户关系管理(Customer Relationship Management，CRM)数据输入至业务预测模型，全量用户的数据和广告主CRM数据可以称为特征数据，通过业务预测模型输出每个用户的第一业务分值，以此确定投放人群的排序，进而对这些人群投放相应的广告，投放广告之后可以继续获取销售线索，在第三个模块中，将第一业务分值输入到业务分值转换模型，进而得到第二业务分值，根据第二业务分值可以对用户购买意向进行排序与分级，后续还有将销售线索加入到业务分值转换模型和业务预测模型中，用于进一步训练业务分值转换模型和业务预测模型。在第四个模块中根据用户排序和分级进行运营引导和销售，最后在第五个模块中反馈真实成交数据。
广告主在广告集客推广中获得了销售线索，该销售线索可以为用户填写的用户基本数据(如年龄、性别、职业等)，与此同时，通过广告数据管理平台(Data-Management Platform，DMP)将真实成交数据通过应用程序编程接口(Application Programming Interface，API) 实时推送到预测模型模块，模型基于线索标识与FTRL框架进行购买意向度预测，并将业务分值结果通过API实时反馈给广告主。在广告集客与销售跟进的过程中，广告主通过广告DMP的API将填写的个人资料和购买等行为数据回传至模型训练模块，形成数据闭环，进行购买意向度预测模型(即业务预测模型)优化。
下面将从业务分值确定装置的角度，对本发明中业务分值的确定方法进行介绍，请参阅图4，本发明实施例中业务分值的确定方法一个实施例包括：
101、获取目标用户的目标特征数据；
本实施例中，业务分值确定装置首先获取目标用户的目标特征数据，其中，目标特征数据包括目标用户的基本数据、语义数据和行为数据中的至少一种，目标用户的基本数据包含但不仅限于目标用户的年龄、性别、地域、学历、职业以及婚恋状态中的至少一种。语义数据包括用户输入或阅读的关键字信息。行为数据包含但不仅限于应用程序的安装，应用程序的活跃指数，QQ空间的操作行为。
102、根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值；
本实施例中，业务分值确定装置将目标用户的目标特征数据输入至业务预测模型，并通过业务预测模型输出目标用户所对应的第一业务分值。
这里的业务预测模型是根据正样本和负样本训练得到的，正样本为销售历史数据中已成交用户对应的样本数据，负样本为销售历史数据中未成交用户对应的样本数据。
正样本和负样本中分别可以包括多种不同的数据，在一种实现方式中，正样本可以包括第一用户基本数据、第一语义数据、第一行为数据；负样本可以包括第二用户基本数据、第二语义数据和第二行为数据。
103、根据业务分值转换模型以及第一业务分值确定第二业务分值，其中，第二业务分值与目标用户的购买意向度呈正相关。
本实施例中，业务分值确定装置再根据业务分值转换模型，将第一业务分值转换为第二业务分值，该业务分值转换模型可以是一种线性变换模型。第二业务分值与目标用户的购买意向度呈正相关，也就是说，第二业务分值越高，意味着目标用户可能会购买该商品的可能性越大，第二业务分值具有较好的解释性，比如50分和60分有着不同的意义，第二业务分值可以根据具体的业务应用场景灵活调整分值范围和疏密程度，便于根据第二业务分值对目标用户进行分级。
本发明实施例中，提供了一种业务分值的确定方法，首先获取目标用户的目标特征数据，然后根据目标特征数据和业务预测模型获取目标用户所对应的第一业务分值，其中，业务预测模型为根据正样本和负样本训练得到的，正样本为销售历史数据中已成交用户对应的样本数据，负样本为销售历史数据中未成交用户对应的样本数据。接下来根据业务分 值转换模型以及第一业务分值确定第二业务分值，其中，第二业务分值与目标用户的购买意向度呈正相关。通过上述方式，对正样本和负样本进行机器学习，建立业务预测模型，进而根据第一业务分值确定出优先人群，从而根据优选人群进行广告投放，以提升集客效率。另外，得到的第二业务分值具有更强的业务意义且区分度明显，更能体现用户购买意向，从而实现优先向购买意向度高的用户进行线索清洗和销售跟进等，提升了方案的实用性和营销效率，节省人力成本。
可选地，在上述图4对应的实施例的基础上，本发明实施例提供的业务分值的确定方法的可选实施例中，通过业务预测模型获取目标特征数据所对应的第一业务分值之前，还可以包括：
获取正样本和负样本；
获取正样本所对应的购买结果以及负样本所对应的购买结果；
对正样本、正样本所对应的购买结果、负样本以及负样本所对应的购买结果进行训练，以得到业务预测模型。
本实施例中，业务分值确定装置在获取第一业务分值之前，需要先得到业务预测模型。一种具体的训练方式是通过逻辑回归(Logistic Regression，LR)进行训练，LR是机器学习中的一种分类模型，由于算法的简单和高效，在实际中应用非常广泛。
LR属于有监督的学习，因此在使用这些算法之前，必须要先收集一批标注好的样本数据作为训练集，其中，样本数据包括正样本和负样本，具体为第一用户基本数据、第一语义数据、第一行为数据、第二用户基本数据、第二语义数据和第二行为数据。第一用户基本数据为已成交用户的个人信息，第一语义数据为已成交用户的关键字信息，第一行为数据为已成交用户的操作信息。第二用户基本数据为未成交用户的个人信息，第二语义数据为未成交用户的关键字信息，第二行为数据为未成交用户的操作信息。有些标注好的样本数据可以从数据库中拿到(例如用户的点击或者购买)，有些标注好的样本数据可以从用户填写的信息中获得(例如性别)，也有一些有些标注好的样本数据是由人工标注的。
下面将通过一个例子来说明如何训练得到业务预测模型。
在实际工作中，LR通常用于解决二分类问题，比如预测一个用户是否点击特定的广告，或预测一个用户是否购买特定的品牌汽车，又或者预测一个用户是否对学习雅思感兴趣等。因此，我们的样本数据可以表示为：
D＝{(x 1,y 1),(x 2,y 2),...,(x N,y N)}；
其中，x i表示第一用户基本数据、第一语义数据以及第一行为数据中的至少一项(或第二用户基本数据、第二语义数据以及第二行为数据中的至少一项)所对应的m维向量，
Figure PCTCN2019071392-appb-000001
y表示购买结果，y等于0的时候表示未购买，y等于1的时候表示已购买，请参阅表1，表1为样本数据的一个示意。
表1
Figure PCTCN2019071392-appb-000002
购买了汽车的样本数据表示正样本，未购买汽车的样本数据表示负样本，假设LR是一个Sigmoid函数，如
Figure PCTCN2019071392-appb-000003
相应的，LR的决策函数就是y *＝1，如果P(y＝1|x)＞threshold，这里的threshold可以取0.5，也可以是其他合理的取值，如果想让正样本的反响更大，那么threshold可以取值小一些。
以表1中第一行的x i为例，预测该用户购买汽车的可能性，该用户的年龄为30，性别为男，常住地为北京，最近7天有摇号，无车，有房，是差旅人士，因此，可以得到x 7＝[30,1,1,1,0,1,1]，分别将x 7中的每个值与对应的似然比参数相乘，在将这7个结果相加后得到P，若P＞threshold，则认为该用户可能会购买车。
训练业务预测模型的过程实际上就是采用最大似然比估计值对LR中的似然比参数θ进行求解的过程，具体地：
L(θ)＝∏P(y|x；θ)＝∏f(θ Tx) y(1-f(θ Tx)) 1-y；
取值log后得到：l(θ)＝∑y log f(θ Tx)+(1-y)log(1-log f(θ Tx))；
将最大似然比估计值转换为最小化损失函数，整个数据集上的平均log损失为
Figure PCTCN2019071392-appb-000004
可以采用梯度下降(Gradient Descent)进行参数最优化，使得损失最小，这样也就能得到最优似然比参数θ，其他凸优化方法有共轭梯度下降，牛顿法。
本发明实施例中，介绍了训练得到业务预测模型的方式，即对正样本和负样本进行逻辑回归训练，可以获取到业务预测模型。通过上述方式，能够提升训练的准确度，同时，逻辑回归训练适用于连续性和类别性的自变量，并且容易使用和解释。
可选地，在上述图4对应的第一个实施例的基础上，本发明实施例提供的业务分值的确定方法第二个可选实施例中，正样本可以包括第一用户基本数据、第一语义数据和第一行为数据，其中，第一用户基本数据为已成交用户的个人信息，第一语义数据为已成交用户的关键字信息，第一行为数据为已成交用户的操作信息；
负样本可以包括第二用户基本数据、第二语义数据和第二行为数据，其中，第二用户基本数据为未成交用户的个人信息，第二语义数据为未成交用户的关键字信息，第二行为数据为未成交用户的操作信息。
本实施例中，业务分值确定装置训练的样本数据可以来源于服务器中的第一用户特征库和/或用户提供的第二用户特征库。
具体地，第一用户特征库可以是腾讯公司提供的用户特征库，该用户特征库的特点是数据量大且全面。而第二用户特征库可以是广告主提供的自有特征数据，比如意向城市和用户历史购买记录等，将第一用户特征库中的数据和第二用户特征库中的数据进行拼接，共同作为正样本和负样本进行训练。这里的拼接过程可以是将具有同样用户标识的数据进行汇总，用户标识可以是手机号或者QQ号等，将更多的数据进行映射和汇总，能够增加数据量的训练维度，从而也提升了样本修正的空间。
在训练业务预测模型之前还需要先获取正样本和负样本。
具体地，用户基本数据中包括基础属性类、社会经济状态类、基于移动位置服务(Location Based Service，LBS)类、设备类、商业兴趣类和垂直行业标签类，语义数据包括语义类，行为数据包括业务行为类。
更具体地，基础属性类包含但不仅限于年龄、性别、地域、学历、职业以及婚恋状态；社会经济状态类包含但不仅限于消于常驻国家、常驻省份以及常驻城市；设备类包含但不仅限于操作系统、运营商、设备类型费能力、购房情况、购车情况、差旅人士、消费记录以及包月服务；LBS类包含但不仅限以及上网场景；商业兴趣类包含但不仅限于浏览行为；垂直行业标签类可以是“汽车_购车潜客_具体到车型”，或者“汽车_有车人群_具体到车型”，或者“汽车_二手车人群_具体到车型”等。
语义类包括用户输入或阅读的关键字信息，比如用户输入“红酒木瓜汤效果怎么样”，首先会对这句话进行分词处理，得到“红酒”、“木瓜”、“汤”、“效果”和“怎么样”，关键字为“红酒木瓜汤”、“红酒木瓜”、“木瓜汤”、“红酒”和“木瓜”，根据关键字映射到对应的话题可以是“美容瘦身”、“美容整形”、“餐饮”和“食品”。
行为数据包含但不仅限于应用程序的安装，应用程序的活跃指数，QQ空间的操作行为。
本发明实施例中，介绍了用于进行训练的正样本和负样本所包含的具体数据，无论是正样本还是负样本都包括了用户基本数据、语义数据和行为数据。通过上述方式，能够获取各种类型的样本数据来进行业务预测模型训练，从而提升样本数据的多样性，以此训练得到的业务预测模型具有更高的准确度。
可选地，在上述图4对应的第一个实施例的基础上，本发明实施例提供的业务分值的确定方法第三个可选实施例中，获取正样本和负样本之前，还可以包括：
获取预处理正样本和预处理负样本；
对预处理正样本进行预处理，以得到正样本；
对预处理负样本进行预处理，以得到负样本，其中，预处理包括去重处理、活跃样本提取和关联样本合并的至少一项。
本实施例中，业务分值确定装置先获取预处理的正样本和预处理的负样本，然后分别对预处理的正样本和预处理的负样本进行预处理。预处理的方式包括但不仅限于去重处理、活跃样本提取和关联样本合并的至少一项。
具体地，下面将分别介绍这三种预处理方式：
第一种去重处理，顾名思义，去重处理就是去掉重复的数据。因为在很多情况下可能会获取到同一个用户对应的多个相同的数据，但是对多个重复的数据进行训练会降低模型的准确度，所以需要剔除这些重复的数据。
第二种活跃样本提取，也就是获取活跃度较高的用户所对应的数据作为样本数据。这是由于样本数据应该具有较强的代表性，用户活跃度高就说明该用户的数据量更大，也就具有更强的可靠性。
第三种关联样本合并，一个用户可能同时绑定了多个账号，比如用户A具有QQ号和微信号，但是这两个账号均与用户A的手机号进行绑定，那么可以确定与该手机号绑定的QQ号和微信号都属于用户A的，因此可以将QQ上的数据与微信上的数据进行合并，得到一份样本数据。
可以理解的是，在实际应用中，多个账号与用户标识之间具有绑定关系，这里的用户标识除了可以是用户的手机号以外，还可是移动设备国际识别码(International Mobile Equipment Identity，IMEI)、QQ号或者媒体访问控制(Media Access Control，MAC)地址。 其中，IMEI又称为国际移动设备标识，是手机的唯一识别号码。MAC地址又称为物理地址或者硬件地址，可以用来定义网络设备的位置，具有全球唯一性。
本发明实施例中，在获取正样本和负样本之前，还需要对正样本和负样本进行预处理，比如进行去重处理、活跃样本提取以及关联样本合并中的至少一种。通过上述方式，在对正样本和负样本进行预处理之后，能够得到更为贴近实际情况的样本数据，从而提升模型训练的准确度。
可选地，在上述图4对应的实施例的基础上，本发明实施例提供的业务分值的确定方法第四个可选实施例中，通过业务预测模型获取目标特征数据所对应的第一业务分值，可以包括：
采用业务预测模型计算第一业务分值：
Figure PCTCN2019071392-appb-000005
其中，所述P(y＝1|x；0)表示所述第一业务分值，所述θ表示似然比参数，所述X表示所述目标特征数据，所述y表示购买结果的概率值，所述T表示转置矩阵。
本实施例中，基于在线机器学习(Follow the regularized Leader，FTRL)训练框架，采用K折交叉验证方法对样本特征数据进行LR训练，得到最优模型，即得到业务预测模型。其中，LR是一种广义的线性回归分析模型，常用来解决分类问题。FTRL训练框架是一种批量处理超大规模的数据集和在线数据流的算法。
通过LR训练得到的决策函数为
Figure PCTCN2019071392-appb-000006
这里的X表示目标特征数据，目标特征数据包含但不仅限于目标用户的用户基本数据、语义数据以及行为数据。P(y＝1|x；0)表示目标特征数据在第一业务分值的购买意向概率，也就是表示这个目标特征数据X在参数θ下购买商品的可能性。其中，对于θ的求解可以采用K折交叉验证方法。
本发明实施例中，提供了一种计算第一业务分值的方式，即利用业务预测模型可以计算得到所需的第一业务分值。通过上述方式，能够提升方案的实用性和可行性。
可选地，在上述图4对应的实施例的基础上，本发明实施例提供的业务分值的确定方法第五个可选实施例中，根据业务分值转换模型以及第一业务分值确定第二业务分值，可以包括：
采用业务分值转换模型计算第二业务分值：
Figure PCTCN2019071392-appb-000007
其中，所述score表示所述第二业务分值，所述average表示所述第一业务分值等于pos_neg_ratio时的预期分值，所述density用于表示所述第二业务分值的疏密程度，所述probability表示所述第一业务分值，所述pos_neg_ratio表示所述正样本与所述负样本之间的比值。
本实施例中，采用业务分值转换模型可以将第一业务分值转换为第二业务分值。具体地，将业务预测模型输出的原始第一业务分值进行转化，得到有业务意义和区分度明显的第二业务分值。；
在利用上述业务分值转换模型计算第二业务分值时，average可以根据具体业务背景进行灵活地调整，一般情况下，average可以设置为期望值分数的平均数，例如期望分值范围为[0,100]，通常情况下可将average设置为50，需要说明的是，期望分值范围是一个可以预先设置的范围，除了[0,100]，还可以是[0,150]，或者[10,160]等范围，此处仅为一个示意，并不应理解为对本发明的限定。
density用于控制第二业务分值的疏密程度，density表示业务场景下第一业务分值翻倍时候的分值，即最终得分的区分度，比如，density为10的情况下，说明第二业务分值每多出10分，表示购买率的可能性翻一倍。density可通过尝试不同取值，观测哪个取值能够得到较好的区分度与业务意义，density一般不超过期望分值的范围。例如：期望分值范围为[0,100]，可以尝试5、10、15、20等density进行观测调优。pos_neg_ratio表示训练样本中，正样本的数量与负样本的数量之间的比值，比如共有10个样本，其中有6个样本是购买了车辆的(即正样本)，4个样本是未购买车辆的(即负样本)，那么pos_neg_ratio为1.5。
以一个场景为例介绍如何通过第二业务分值预测购买率。假设density为10分，average为50分的时候表示正负样本之间的比值pos_neg_ratio为2.3％，此时，得到第二业务分值为60分的时候，就意味着用户的购买率为4.6％。
又假设density为5分，average为50分的时候表示正负样本之间的比值pos_neg_ratio为2.3％，此时，得到第二业务分值为60分的时候，就意味着用户的购买率为9.2％。
为了便于理解，下面将以一个示例来说明如何推导得到分值转换模型，具体地，假设用户购买的概率表示为probability，该probability表示第一业务分值，不购买的概率表示为(1-probability)，则
Figure PCTCN2019071392-appb-000008
probability越高则score越高，因此将score表示为log(pos_neg_ratio)的线性表达式：
score＝A+B*log(pos_neg_ratios)；
其中，A和B的值需要通过两个假设的分值代入计算后得到，这两个假设分别为：
第一个假设，给某个特定的正负样本比值pos_neg_ratio设定为一个特定的预期分值average；
第二个假设，确定比率翻倍的分数density，当density等于10，意味着每个增加10分pos_neg_ratio可以翻倍。
根据上述的假设，可以得到如下两个等式：
average＝A+B+log(pos_neg_ratio)；
average+density＝A+B+log(2*pos_neg_ratio)；
由此，求解得到
A＝average-B*log(pos_neg_ratio)；
B＝density/log(2)；
最后得到业务分值转换模型：
Figure PCTCN2019071392-appb-000009
可以理解的是，根据业务需要，期望分值范围可以为0至100，而average可以取50，density可以取值10。
其次，本发明实施例中，提供了一种计算第二业务分值的方式，即利用业务分值转换模型可以计算得到所需的第二业务分值。通过上述方式，能够提升方案的实用性和可行性。
可选地，在上述图4以及图4对应的第一个至第五个实施例中任一项的基础上，目标用户可以是需要进行业务分析的任一用户，即针对每个目标用户都可以执行前述101-103的步骤，从而业务分值确定装置可以得到每个目标用户的第二业务分值，然后，根据第二业务分值的大小对不同目标用户进行排序，排序结果可以反映出不同目标用户购买意向度的大小顺序。
具体地，假设当前需要对100名用户进行业务分析，首先分别将每位用户作为目标用户，从而计算每位用户所对应的第二业务分值，若有20名用户的第二业务分值达到阈值，那么这20名用户就是高潜用户，可以作为高潜线索。然后在对这20名用户的第二业务分值进行排序，排序结果可以反映出不同目标用户购买意向度的大小顺序，客服人员可以依据第二业务分值从高到低的顺序进行线索清洗，即依据购买意向度从大到小的顺序对目标用户进行线索清洗和销售跟进等。
本发明实施例中，在得到多个用户的业务分值之后，可以按照从第二业务分值从高到低的顺序对用户进行排序。通过上述方式，广告主在获取各个用户的第二业务分值后，可以由客服人员依据第二业务分值高低进行线索清洗，从而实现对高购买意向度用户的高效跟进，以此提升销售的达成率。
下面对本发明中的业务分值确定装置进行详细描述，请参阅图5，图5为本发明实施例中业务分值确定装置一个实施例示意图，业务分值确定装置20包括：
获取模块201，用于获取目标用户的目标特征数据；
所述获取模块201，用于根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值，其中，所述业务预测模型为根据正样本和负样本训练得到的，所述正样本为销售历史数据中已成交用户对应的样本数据，所述负样本为销售历史数据中未成交用户对应的样本数据；
确定模块202，用于根据业务分值转换模型以及所述获取模块201获取的所述第一业务分值确定第二业务分值，其中，所述第二业务分值与所述目标用户的购买意向度呈正相关。
本实施例中，获取模块201获取目标用户的目标特征数据，所述获取模块201将所述目标特征数据输入到业务预测模型，从而输出所述目标用户所对应的第一业务分值，确定模块202通过业务分值转换模型将所述获取模块201获取的所述第一业务分值转换为第二业务分值。
本发明实施例中，提供了一种业务分值确定装置，首先获取目标用户的目标特征数据，然后根据目标特征数据和业务预测模型获取目标用户所对应的第一业务分值，其中，业务预测模型为根据正样本和负样本训练得到的，正样本为销售历史数据中已成交用户对应的样本数据，负样本为销售历史数据中未成交用户对应的样本数据。接下来根据业务分值转换模型以及第一业务分值确定第二业务分值，其中，第二业务分值与目标用户的购买意向度呈正相关。通过上述方式，对正样本和负样本进行机器学习，建立业务预测模型，进而根据第一业务分值确定出优先人群，从而根据优选人群进行广告投放，以提升集客效率。另外，得到的第二业务分值具有更强的业务意义且区分度明显，更能体现用户购买意向，从而实现优先向购买意向度高的用户进行线索清洗和销售跟进等，提升了方案的实用性和营销效率，节省人力成本。
可选地，在上述图5所对应的实施例的基础上，请参阅图6，本发明实施例提供的业务分值确定装置20的另一实施例中，所述业务分值确定装置20还包括训练模块203；
所述获取模块201，还用于根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值之前，获取所述正样本和所述负样本；
获取所述正样本所对应的购买结果以及所述负样本所对应的购买结果；
所述训练模块203，用于对所述获取模块201获取的所述正样本、所述正样本所对应的购买结果、所述负样本以及所述负样本所对应的购买结果进行训练，以得到所述业务预测模型。
其次，本发明实施例中，介绍了训练得到业务预测模型的方式，即对正样本和负样本进行逻辑回归训练，可以获取到业务预测模型。通过上述方式，能够提升训练的准确度，同时，逻辑回归训练适用于连续性和类别性的自变量，并且容易使用和解释。
可选地，在上述图6所对应的实施例的基础上，本发明实施例提供的业务分值确定装置20的另一实施例中，
所述正样本包括所述第一用户基本数据、所述第一语义数据和所述第一行为数据，其中，所述第一用户基本数据为已成交用户的个人信息，所述第一语义数据为所述已成交用户的关键字信息，所述第一行为数据为所述已成交用户的操作信息；
所述负样本包括所述第二用户基本数据、所述第二语义数据和所述第二行为数据，其中，所述第二用户基本数据为未成交用户的个人信息，所述第二语义数据为所述未成交用户的关键字信息，所述第二行为数据为所述未成交用户的操作信息。
本发明实施例中，介绍了用于进行训练的正样本和负样本所包含的具体数据，无论是正样本还是负样本都包括了用户基本数据、语义数据和行为数据。通过上述方式，能够获取各种类型的样本数据来进行业务预测模型训练，从而提升样本的多样性，以此训练得到的业务预测模型具有更高的准确度。
可选地，在上述图5所对应的实施例的基础上，本发明实施例提供的业务分值确定装置20的另一实施例中，
所述获取模块201，具体用于采用所述业务预测模型计算所述第一业务分值：
Figure PCTCN2019071392-appb-000010
其中，所述P(y＝1|x；0)表示所述第一业务分值，所述θ表示似然比参数，所述X表示所述目标特征数据，所述y表示购买结果的概率值，所述T表示转置矩阵。
其次，本发明实施例中，提供了一种计算第一业务分值的方式，即利用业务预测模型可以计算得到所需的第一业务分值。通过上述方式，能够提升方案的实用性和可行性。
可选地，在上述图5所对应的实施例的基础上，本发明实施例提供的业务分值确定装置20的另一实施例中，
所述确定模块202，用具体于采用所述业务分值转换模型计算所述第二业务分值：
Figure PCTCN2019071392-appb-000011
其中，所述score表示所述第二业务分值，所述average表示所述第一业务分值等于pos_neg_ratio时的预期分值，所述density用于表示所述第二业务分值的疏密程度，所述probability表示所述第一业务分值，所述pos_neg_ratio表示所述正样本与所述负样本之间的比值。
本发明实施例中，提供了一种计算第二业务分值的方式，即利用业务分值转换模型可以计算得到所需的第二业务分值。通过上述方式，能够提升方案的实用性和可行性。
进一步地，本发明实施例中，在得到多个用户的第二业务分值之后，可以按照第二业务分值从高到低的顺序对用户进行排序。通过上述方式，广告主在获取各个用户的第二业务分值后，可以由客服人员依据第二业务分值高低进行线索清洗，从而实现对高购买意向度用户的高效跟进，以此提升销售的达成率。
本发明实施例还提供了另一种业务分值确定装置，如图7所示，为了便于说明，仅示出了与本发明实施例相关的部分，具体技术细节未揭示的，请参照本发明实施例方法部分。该终端可以为包括手机、平板电脑、个人数字助理(Personal Digital Assistant，PDA)、销售终端(Point of Sales，POS)、车载电脑等任意终端设备，以终端为手机为例：
图7示出的是与本发明实施例提供的终端相关的手机的部分结构的框图。参考图7，手机包括：射频(Radio Frequency，RF)电路310、存储器320、输入单元330、显示单元340、传感器350、音频电路360、无线保真(wireless fidelity，WiFi)模块370、处理器380、以及电源390等部件。本领域技术人员可以理解，图7中示出的手机结构并不构成对手机的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。
下面结合图7对手机的各个构成部件进行具体的介绍：
RF电路310可用于收发信息或通话过程中，信号的接收和发送，特别地，将基站的下行信息接收后，给处理器380处理；另外，将设计上行的数据发送给基站。通常，RF电路310包括但不限于天线、至少一个放大器、收发信机、耦合器、低噪声放大器(Low Noise Amplifier，LNA)、双工器等。此外，RF电路310还可以通过无线通信与网络和其他设备通信。上述无线通信可以使用任一通信标准或协议，包括但不限于全球移动通讯系统(Global System of Mobile communication，GSM)、通用分组无线服务(General Packet Radio Service，GPRS)、码分多址(Code Division Multiple Access，CDMA)、宽带码分多址(Wideband Code Division Multiple Access,WCDMA)、长期演进(Long Term Evolution，LTE)、电子邮件、短消息服务(Short Messaging Service，SMS)等。
存储器320可用于存储软件程序以及模块，处理器380通过运行存储在存储器320的软件程序以及模块，从而执行手机的各种功能应用以及数据处理。存储器320可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需的应用程序(比如声音播放功能、图像播放功能等)等；存储数据区可存储根据手机的使用所创建的数据(比如音频数据、电话本等)等。此外，存储器320可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他易失性固态存储器件。
输入单元330可用于接收输入的数字或字符信息，以及产生与手机的用户设置以及功能控制有关的键信号输入。具体地，输入单元330可包括触控面板331以及其他输入设备332。触控面板331，也称为触摸屏，可收集用户在其上或附近的触摸操作(比如用户使用手指、触笔等任何适合的物体或附件在触控面板331上或在触控面板331附近的操作)，并根据预先设定的程式驱动相应的连接装置。可选的，触控面板331可包括触摸检测装置和触摸控制器两个部分。其中，触摸检测装置检测用户的触摸方位，并检测触摸操作带来的信号，将信号传送给触摸控制器；触摸控制器从触摸检测装置上接收触摸信息，并将它转换成触点坐标，再送给处理器380，并能接收处理器380发来的命令并加以执行。此外， 可以采用电阻式、电容式、红外线以及表面声波等多种类型实现触控面板331。除了触控面板331，输入单元330还可以包括其他输入设备332。具体地，其他输入设备332可以包括但不限于物理键盘、功能键(比如音量控制按键、开关按键等)、轨迹球、鼠标、操作杆等中的一种或多种。
显示单元340可用于显示由用户输入的信息或提供给用户的信息以及手机的各种菜单。显示单元340可包括显示面板341，可选的，可以采用液晶显示器(Liquid Crystal Display，LCD)、有机发光二极管(Organic Light-Emitting Diode，OLED)等形式来配置显示面板341。进一步的，触控面板331可覆盖显示面板341，当触控面板331检测到在其上或附近的触摸操作后，传送给处理器380以确定触摸事件的类型，随后处理器380根据触摸事件的类型在显示面板341上提供相应的视觉输出。虽然在图7中，触控面板331与显示面板341是作为两个独立的部件来实现手机的输入和输入功能，但是在某些实施例中，可以将触控面板331与显示面板341集成而实现手机的输入和输出功能。
手机还可包括至少一种传感器350，比如光传感器、运动传感器以及其他传感器。具体地，光传感器可包括环境光传感器及接近传感器，其中，环境光传感器可根据环境光线的明暗来调节显示面板341的亮度，接近传感器可在手机移动到耳边时，关闭显示面板341和/或背光。作为运动传感器的一种，加速计传感器可检测各个方向上(一般为三轴)加速度的大小，静止时可检测出重力的大小及方向，可用于识别手机姿态的应用(比如横竖屏切换、相关游戏、磁力计姿态校准)、振动识别相关功能(比如计步器、敲击)等；至于手机还可配置的陀螺仪、气压计、湿度计、温度计、红外线传感器等其他传感器，在此不再赘述。
音频电路360、扬声器361，传声器362可提供用户与手机之间的音频接口。音频电路360可将接收到的音频数据转换后的电信号，传输到扬声器361，由扬声器361转换为声音信号输出；另一方面，传声器362将收集的声音信号转换为电信号，由音频电路360接收后转换为音频数据，再将音频数据输出处理器380处理后，经RF电路310以发送给比如另一手机，或者将音频数据输出至存储器320以便进一步处理。
WiFi属于短距离无线传输技术，手机通过WiFi模块370可以帮助用户收发电子邮件、浏览网页和访问流式媒体等，它为用户提供了无线的宽带互联网访问。虽然图7示出了WiFi模块370，但是可以理解的是，其并不属于手机的必须构成，完全可以根据需要在不改变发明的本质的范围内而省略。
处理器380是手机的控制中心，利用各种接口和线路连接整个手机的各个部分，通过运行或执行存储在存储器320内的软件程序和/或模块，以及调用存储在存储器320内的数据，执行手机的各种功能和处理数据，从而对手机进行整体监控。可选的，处理器380可包括一个或多个处理单元；可选的，处理器380可集成应用处理器和调制解调处理器，其中，应用处理器主要处理操作系统、用户界面和应用程序等，调制解调处理器主要处理无线通信。可以理解的是，上述调制解调处理器也可以不集成到处理器380中。
手机还包括给各个部件供电的电源390(比如电池)，可选的，电源可以通过电源管理系统与处理器380逻辑相连，从而通过电源管理系统实现管理充电、放电、以及功耗管理等功能。
尽管未示出，手机还可以包括摄像头、蓝牙模块等，在此不再赘述。
在本发明实施例中，该终端所包括的处理器380还具有以下功能：
获取目标用户的目标特征数据；
根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值，其中，所述业务预测模型为根据正样本和负样本训练得到的，所述正样本为销售历史数据中已成交用户对应的样本数据，所述负样本为销售历史数据中未成交用户对应的样本数据；
根据业务分值转换模型以及所述第一业务分值确定第二业务分值，其中，所述第二业务分值与所述目标用户的购买意向度呈正相关。
可选地，处理器380还用于执行如下步骤：
获取所述正样本和所述负样本；
获取所述正样本所对应的购买结果以及所述负样本所对应的购买结果；
对所述正样本、所述正样本所对应的购买结果、所述负样本以及所述负样本所对应的购买结果进行训练，以得到所述业务预测模型。
可选地，处理器380还用于执行如下步骤：
获取待处理正样本和待处理负样本；
对所述待处理正样本进行预处理，以得到所述正样本；
对所述待处理负样本进行预处理，以得到所述负样本，其中，所述预处理包括去重处理、活跃样本提取和关联样本合并的至少一项。
可选地，处理器380具体用于执行如下步骤：
采用所述业务预测模型计算所述第一业务分值：
Figure PCTCN2019071392-appb-000012
其中，所述P(y＝1|x；0)表示所述第一业务分值，所述θ表示似然比参数，所述X表示所述目标特征数据，所述y表示购买结果的概率值，所述T表示转置矩阵。
可选地，处理器380具体用于执行如下步骤：
采用所述业务分值转换模型计算所述第二业务分值：
Figure PCTCN2019071392-appb-000013
其中，所述score表示所述第二业务分值，所述average表示所述第一业务分值等于pos_neg_ratio时的预期分值，所述density用于表示所述第二业务分值的疏密程度，所述probability表示所述第一业务分值，所述pos_neg_ratio表示所述正样本与所述负样本之间的比值。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统，装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本发明所提供的几个实施例中，应该理解到，所揭露的系统，装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本发明各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本发明的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本发明各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(read-only memory，ROM)、随机存取存储器(random access memory，RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，以上实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的精神和范围。

Claims (11)

  1. 一种业务分值的确定方法，所述方法应用于业务分值确定装置，包括：
    获取目标用户的目标特征数据；
    根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值，其中，所述业务预测模型为根据正样本和负样本训练得到的，所述正样本为销售历史数据中已成交用户对应的样本数据，所述负样本为销售历史数据中未成交用户对应的样本数据；
    根据业务分值转换模型以及所述第一业务分值确定第二业务分值，其中，所述第二业务分值与所述目标用户的购买意向度呈正相关。
  2. 根据权利要求1所述的方法，所述根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值之前，所述方法还包括：
    获取所述正样本和所述负样本；
    获取所述正样本所对应的购买结果以及所述负样本所对应的购买结果；
    对所述正样本、所述正样本所对应的购买结果、所述负样本以及所述负样本所对应的购买结果进行训练，以得到所述业务预测模型。
  3. 根据权利要求2所述的方法，
    所述正样本包括第一用户基本数据、第一语义数据和第一行为数据，其中，所述第一用户基本数据包括已成交用户的个人信息，所述第一语义数据包括所述已成交用户的关键字信息，所述第一行为数据包括所述已成交用户的操作信息；
    所述负样本包括第二用户基本数据、第二语义数据和第二行为数据，其中，所述第二用户基本数据包括未成交用户的个人信息，所述第二语义数据包括所述未成交用户的关键字信息，所述第二行为数据包括所述未成交用户的操作信息。
  4. 根据权利要求1所述的方法，所述根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值，包括：
    采用所述业务预测模型计算所述第一业务分值：
    Figure PCTCN2019071392-appb-100001
    其中，所述P(y＝1|x；0)表示所述第一业务分值，所述θ表示似然比参数，所述X表示所述目标特征数据，所述y表示购买结果的概率值，所述T表示转置矩阵。
  5. 根据权利要求1所述的方法，所述根据业务分值转换模型以及所述第一业务分值确定第二业务分值，包括：
    采用所述业务分值转换模型计算所述第二业务分值：
    Figure PCTCN2019071392-appb-100002
    其中，所述score表示所述第二业务分值，所述average表示所述第一业务分值等于pos_neg_ratio时的预期分值，所述density用于表示所述第二业务分值的疏密程度，所述probability表示所述第一业务分值，所述pos_neg_ratio表示所述正样本与所述负样本之间的比值。
  6. 一种业务分值确定装置，包括：
    获取模块，用于获取目标用户的目标特征数据；
    所述获取模块，用于根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值，其中，所述业务预测模型为根据正样本和负样本训练得到的，所述正样本为销售历史数据中已成交用户对应的样本数据，所述负样本为销售历史数据中未成交用户对应的样本数据；
    确定模块，用于根据业务分值转换模型以及所述获取模块获取的所述第一业务分值确定第二业务分值，其中，所述第二业务分值与所述目标用户的购买意向度呈正相关。
  7. 根据权利要求6所述的业务分值确定装置，所述业务分值确定装置还包括训练模块；
    所述获取模块，还用于根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值之前，获取所述正样本和所述负样本；
    获取所述正样本所对应的购买结果以及所述负样本所对应的购买结果；
    所述训练模块，用于对所述获取模块获取的所述正样本、所述正样本所对应的购买结果、所述负样本以及所述负样本所对应的购买结果进行训练，以得到所述业务预测模型。
  8. 根据权利要求7所述的业务分值确定装置，
    所述获取模块，用于获取第一用户基本数据、所述第一语义数据和第一行为数据，其中，所述第一用户基本数据包括已成交用户的个人信息，所述第一语义数据包括所述已成交用户的关键字信息，所述第一行为数据包括所述已成交用户的操作信息；
    获取第二用户基本数据、第二语义数据和第二行为数据，其中，所述第二用户基本数据包括未成交用户的个人信息，所述第二语义数据包括所述未成交用户的关键字信息，所述第二行为数据包括所述未成交用户的操作信息。
  9. 一种业务分值确定装置，包括：存储器、收发器、处理器以及总线系统；
    其中，所述存储器用于存储程序；
    所述处理器用于执行所述存储器中的程序，包括如下步骤：
    获取目标用户的目标特征数据；
    根据所述目标特征数据和业务预测模型获取所述目标用户所对应的第一业务分值，其中，所述业务预测模型为根据正样本和负样本训练得到的，所述正样本为销售历史数据中已成交用户对应的样本数据，所述负样本为销售历史数据中未成交用户对应的样本数据；
    根据业务分值转换模型以及所述第一业务分值确定第二业务分值，其中，所述第二业务分值与所述目标用户的购买意向度呈正相关；
    所述总线系统用于连接所述存储器以及所述处理器，以使所述存储器以及所述处理器进行通信。
  10. 一种计算机可读存储介质，包括指令，当其在计算机上运行时，使得计算机执行如权利要求1至5中任一项所述的方法。
  11. 一种计算机程序产品，包括指令，当其在计算机上运行时，使得计算机执行如权利要求1-5中任一项所述的方法。
PCT/CN2019/071392 2018-01-12 2019-01-11 一种业务分值的确定方法、装置及存储介质 WO2019137485A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810033512.6 2018-01-12
CN201810033512.6A CN110033294A (zh) 2018-01-12 2018-01-12 一种业务分值的确定方法、业务分值确定装置及介质

Publications (1)

Publication Number Publication Date
WO2019137485A1 true WO2019137485A1 (zh) 2019-07-18

Family

ID=67218899

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/071392 WO2019137485A1 (zh) 2018-01-12 2019-01-11 一种业务分值的确定方法、装置及存储介质

Country Status (2)

Country Link
CN (1) CN110033294A (zh)
WO (1) WO2019137485A1 (zh)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110458220B (zh) * 2019-07-31 2024-04-12 腾讯科技（深圳）有限公司 人群定向方法、装置、服务器及存储介质
CN111046274B (zh) * 2019-11-08 2023-07-25 微梦创科网络科技（中国）有限公司 一种基于实时博文的信息推送方法及装置
CN111415195B (zh) * 2020-03-17 2024-04-16 京东科技控股股份有限公司 一种广告推荐方法、装置、终端设备及存储介质
CN111507751A (zh) * 2020-03-26 2020-08-07 北京睿科伦智能科技有限公司 一种基于沟通数据的线索打分方法
CN111612519B (zh) * 2020-04-13 2023-11-21 广发证券股份有限公司 一种识别金融产品潜在客户的方法、装置及存储介质
CN112053184B (zh) * 2020-08-20 2024-01-30 腾讯科技（深圳）有限公司 推广信息投放方法、装置、电子设备以及存储介质
CN113256339B (zh) * 2021-06-04 2023-08-29 脸萌有限公司 资源投放的方法、装置、存储介质及电子设备
CN113190599B (zh) * 2021-06-30 2021-09-28 平安科技（深圳）有限公司 应用用户行为数据的处理方法、装置、设备及存储介质

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100306123A1 (en) * 2009-05-31 2010-12-02 International Business Machines Corporation Information retrieval method, user comment processing method, and systems thereof
CN105868847A (zh) * 2016-03-24 2016-08-17 车智互联（北京）科技有限公司 一种购物行为的预测方法及装置
CN107093084A (zh) * 2016-08-01 2017-08-25 北京小度信息科技有限公司 潜在用户预测转化方法及装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9584836B2 (en) * 2014-09-03 2017-02-28 International Business Machines Corporation Mobility enhanced advertising on internet protocol television
CN105160008B (zh) * 2015-09-21 2020-03-31 合一网络技术(北京)有限公司 一种定位推荐用户的方法及装置
CN107341272A (zh) * 2017-08-25 2017-11-10 北京奇艺世纪科技有限公司 一种推送方法、装置和电子设备

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100306123A1 (en) * 2009-05-31 2010-12-02 International Business Machines Corporation Information retrieval method, user comment processing method, and systems thereof
CN105868847A (zh) * 2016-03-24 2016-08-17 车智互联（北京）科技有限公司 一种购物行为的预测方法及装置
CN107093084A (zh) * 2016-08-01 2017-08-25 北京小度信息科技有限公司 潜在用户预测转化方法及装置

Also Published As

Publication number Publication date
CN110033294A (zh) 2019-07-19

Similar Documents

Publication Publication Date Title
WO2019137485A1 (zh) 一种业务分值的确定方法、装置及存储介质
US11574139B2 (en) Information pushing method, storage medium and server
WO2018059176A1 (zh) 定向标签生成方法、装置及存储介质
CA2700030C (en) Touchpoint customization system
US20190287155A1 (en) Landing page providing server and method of providing customized landing page
CN112115363A (zh) 一种推荐方法、计算设备及存储介质
WO2021120875A1 (zh) 搜索方法、装置、终端设备及存储介质
US20190378171A1 (en) Targeted advertisement system
CN107679561A (zh) 图像处理方法和装置、系统、计算机设备
WO2022247666A1 (zh) 一种内容的处理方法、装置、计算机设备和存储介质
US20230074212A1 (en) Systems and methods for orienting webpage content based on user attention
WO2022247722A1 (zh) 用于投放广告的方法及装置
JP2023507043A (ja) データ処理方法、装置、機器、記憶媒体及びコンピュータプログラム
CN115271931A (zh) 一种信用卡产品的推荐方法、装置、电子设备和介质
CN111787042B (zh) 用于推送信息的方法和装置
CN112257756A (zh) 模型训练数据的选取方法及装置
CN112115710A (zh) 一种行业信息识别方法及装置
US20210312502A1 (en) System and method for generating customized advertisement
KR102652416B1 (ko) 고속 유사도 연산 기반의 상품에 대한 정보를 제공하기 위한 방법 및 이를 위한 장치
CN114547242A (zh) 一种问卷调研的方法、装置、电子设备及可读存储介质
CN116205686A (zh) 一种多媒体资源推荐的方法、装置、设备和存储介质
CN115239356A (zh) 一种推荐内容的管理方法以及相关装置
US20140200971A1 (en) System and method for matching informative content to a multimedia content element based on concept recognition of the multimedia content
CN112860980A (zh) 内容展现、推荐的处理方法、装置、存储介质及电子设备
US20190065607A1 (en) Automated application analytics

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19738710

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19738710

Country of ref document: EP

Kind code of ref document: A1