CN114580790A - Life cycle stage prediction and model training method, device, medium and equipment - Google Patents

Life cycle stage prediction and model training method, device, medium and equipment Download PDF

Info

Publication number
CN114580790A
CN114580790A CN202210376202.0A CN202210376202A CN114580790A CN 114580790 A CN114580790 A CN 114580790A CN 202210376202 A CN202210376202 A CN 202210376202A CN 114580790 A CN114580790 A CN 114580790A
Authority
CN
China
Prior art keywords
media object
sample
value
predicted
stage
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210376202.0A
Other languages
Chinese (zh)
Inventor
李锦添
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Netease Cloud Music Technology Co Ltd
Original Assignee
Hangzhou Netease Cloud Music Technology Co Ltd
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 Hangzhou Netease Cloud Music Technology Co Ltd filed Critical Hangzhou Netease Cloud Music Technology Co Ltd
Priority to CN202210376202.0A priority Critical patent/CN114580790A/en
Publication of CN114580790A publication Critical patent/CN114580790A/en
Pending legal-status Critical Current

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
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Development Economics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Game Theory and Decision Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The embodiment of the disclosure relates to a method and a device for predicting a life cycle phase of a media object and training a phase prediction model, a storage medium and electronic equipment, and relates to the technical field of computers. The method comprises the following steps: acquiring a characteristic data set associated with a media object to be predicted and a target user; inputting the characteristic data set into a pre-trained stage prediction model to obtain a first prediction value of the media object to be predicted; determining a user weight of a target user; determining a second predicted value of the media object to be predicted according to the first predicted value and the user weight; and determining the current life cycle stage of the media object to be predicted according to the second predicted value. The present disclosure promotes prediction accuracy of a lifecycle stage of a media object.

Description

Life cycle stage prediction and model training method, device, medium and equipment
Technical Field
Embodiments of the present disclosure relate to the field of computer technologies, and in particular, to a method, an apparatus, a medium, and a device for predicting a life cycle stage of a media object and training a model.
Background
This section is intended to provide a background or context to the embodiments of the disclosure recited in the claims and the description herein is not admitted to be prior art by inclusion in this section.
With the development of internet technology, media services based on internet technology are processing a rapid development stage, and media services may transmit data of media objects based on internet technology, where the media objects may be songs, audio, or video.
Typically, the media object platform may provide rich resources for the user to play the media object. As with other user-oriented products, the media object may also undergo different lifecycle stages after coming online, e.g., the lifecycle stages of the media object may include a silent period, a growth period, and a decay period.
Disclosure of Invention
However, in the conventional media object life cycle phase prediction technology, the accuracy of predicting the life cycle phase of the media object is low.
Therefore, a method for predicting the life cycle phase of a media object is needed, which can predict the life cycle phase of the media object more accurately.
In this context, embodiments of the present disclosure are intended to provide a media object lifecycle stage prediction and stage prediction model training method, apparatus, computer-readable storage medium, and electronic device.
According to a first aspect of the disclosed embodiments, there is provided a media object lifecycle stage prediction method, including:
acquiring a characteristic data set associated with a media object to be predicted and a target user;
inputting the characteristic data set into a pre-trained stage prediction model to obtain a first prediction value of the media object to be predicted;
determining a user weight of the target user;
determining a second predicted value of the media object to be predicted according to the first predicted value and the user weight;
and determining the current life cycle stage of the media object to be predicted according to the second predicted value.
According to a second aspect of the embodiments of the present disclosure, there is provided a phase prediction model training method, including:
collecting a sample characteristic data set of a sample media object associated with a target user;
determining a life cycle stage tag value of the sample media object according to life cycle information of the sample media object and the sample characteristic data set;
inputting the sample characteristic data set into a stage prediction model to be trained to obtain a sample first prediction value of the sample media object;
determining a loss function value according to the life cycle stage label value and the sample first predicted value;
and updating the model parameters of the stage prediction model according to the loss function values.
According to a third aspect of the disclosed embodiments, there is provided a media object lifecycle stage prediction apparatus, the apparatus comprising:
a first obtaining module configured to obtain a feature data set of a media object to be predicted associated with a target user;
the first prediction module is configured to input the feature data set into a pre-trained stage prediction model to obtain a first prediction value of the media object to be predicted;
a first determination module configured to determine a user weight of the target user;
a second determination module configured to determine a second predicted value of the media object to be predicted according to the first predicted value and the user weight;
and the third determination module is configured to determine the current life cycle stage of the media object to be predicted according to the second predicted value.
According to a fourth aspect of the embodiments of the present disclosure, there is provided a stage prediction model training apparatus including:
an acquisition module configured to acquire a sample feature dataset of a sample media object associated with a target user;
a fifth determining module configured to determine a lifecycle stage tag value for the sample media object as a function of the lifecycle information for the sample media object and the sample feature dataset;
the model training module is configured to input the sample characteristic data set into a stage prediction model to be trained to obtain a sample first prediction value of the sample media object;
a sixth determining module configured to determine a loss function value from the lifecycle stage tag value and a sample first prediction value;
a model update module configured to update model parameters of the phase prediction model according to the loss function values.
According to a fifth aspect of embodiments of the present disclosure, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of the first or second aspect described above.
According to a sixth aspect of the disclosed embodiments, there is provided an electronic device comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the method of the first or second aspect described above via execution of the executable instructions.
Drawings
The above and other objects, features and advantages of exemplary embodiments of the present disclosure will become readily apparent from the following detailed description read in conjunction with the accompanying drawings. Several embodiments of the present disclosure are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which:
FIG. 1 shows a schematic diagram of a media object lifecycle stage prediction system architecture, according to an embodiment of the present disclosure;
FIG. 2 illustrates a flow chart of a method for media object lifecycle stage prediction, in accordance with an embodiment of the present disclosure;
FIG. 3 illustrates a flow chart of a method of phase prediction model training in accordance with an embodiment of the present disclosure;
FIG. 4 illustrates a graph of temporal information for different lifecycle stages of a sample media object, according to an embodiment of the disclosure;
FIG. 5 is a diagram illustrating a relationship between the number of played media objects and a sample weight of a target user according to an embodiment of the disclosure;
FIG. 6 shows a schematic diagram of a media object lifecycle stage prediction apparatus according to an embodiment of the present disclosure;
FIG. 7 illustrates a schematic diagram of a phase prediction model training apparatus in accordance with an embodiment of the present disclosure;
FIG. 8 shows a block diagram of an electronic device according to an embodiment of the disclosure.
In the drawings, the same or corresponding reference numerals indicate the same or corresponding parts.
Detailed Description
The principles and spirit of the present disclosure will be described with reference to a number of exemplary embodiments. It is understood that these embodiments are presented merely to enable those skilled in the art to better understand and to practice the disclosure, and are not intended to limit the scope of the disclosure in any way. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
As will be appreciated by one skilled in the art, embodiments of the present disclosure may be embodied as a system, apparatus, device, method, or computer program product. Accordingly, the present disclosure may be embodied in the form of: entirely hardware, entirely software (including firmware, resident software, micro-code, etc.), or a combination of hardware and software.
According to the embodiment of the disclosure, a method, a device, a computer-readable storage medium and an electronic device for media object life cycle stage prediction and stage prediction model training are provided.
In this document, any number of elements in the drawings is intended to be illustrative and not restrictive, and any nomenclature is used for distinction only and not for any restrictive meaning.
The principles and spirit of the present disclosure are explained in detail below with reference to several representative embodiments of the present disclosure.
Summary of The Invention
The inventor finds that, for a media object, the media object life cycle stage prediction technology provided by the prior art cannot accurately predict the life cycle stage of the media object currently. For example, a prediction technique provided by the prior art may obtain an online time of a media object, determine the current number of days of online of the media object according to the online time, and distinguish a life cycle stage of music according to the number of days of online. The media object life cycle stage prediction technology only takes the online time of a media object as a reference condition to determine the life cycle stage of the media object, the reference condition is single, and the accuracy of the determined life cycle stage of the media object is low; another prediction technique provided by the prior art can monitor and acquire the playing heat of the media object, and determine whether the media object reaches a certain life cycle stage according to the playing heat and an empirical value. The process usually depends on a large number of operators to mark the playing heat of the media object, which consumes manpower; meanwhile, as the playing heat of the media object generates random fluctuation along with the time, the random fluctuation is gradually stabilized only after the time is long enough, and the playing heat can not be utilized to make accurate prediction on the life cycle stage of the media object at the online initial stage of the media object.
In view of the above, the basic idea of the present disclosure is: the method, the device, the computer readable storage medium and the electronic equipment can be used for pre-training a phase prediction model for predicting the life cycle phase of a media object, and for the media object to be predicted, a feature data set associated with the media object to be predicted and a target user can be obtained; inputting the characteristic data set into a pre-trained stage prediction model to obtain a first prediction value of the media object to be predicted; determining a user weight of a target user; determining a second predicted value of the media object to be predicted according to the first predicted value and the user weight; and determining the current life cycle stage of the media object to be predicted according to the second predicted value. The feature data set associated with the media object by the user and the user weight in the life cycle phase prediction of the media object by the user can be combined, so that the life cycle phase of the media object is estimated more accurately.
It should be noted that, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for presentation, analyzed data, etc.) referred to in the present disclosure are information and data authorized by the user or sufficiently authorized by each party; and the embodiments/examples provided in the present disclosure may be combined with each other.
Having described the general principles of the present disclosure, various non-limiting embodiments of the present disclosure are described in detail below.
Application scene overview
It should be noted that the following application scenarios are merely illustrated to facilitate understanding of the spirit and principles of the present disclosure, and embodiments of the present disclosure are not limited in this respect. Rather, embodiments of the present disclosure may be applied to any scenario where applicable.
The present disclosure provides an architecture diagram for a media object lifecycle stage prediction system. As shown in fig. 1, the system architecture 100 may include: user terminal 110 and server 120, user terminal 110 may be a terminal device of a media object platform operator, and terminal device 110 may be a smartphone, a tablet computer, a personal computer, or an intelligent wearable device. Server 120 may be a server of a media object platform. User terminal 110 may establish a network connection with server 120 for media object lifecycle stage prediction.
The user terminal 110 may send a life cycle prediction request to the server 120, and the server 120 may receive the life cycle prediction request, obtain a to-be-predicted media object identifier in the life cycle prediction request, and obtain a feature data set associated with the to-be-predicted media object and the target user; inputting the characteristic data set into a pre-trained stage prediction model to obtain a first prediction value of the media object to be predicted; determining a user weight of a target user; determining a second predicted value of the media object to be predicted according to the first predicted value and the user weight; and determining the current life cycle stage of the media object to be predicted according to the second predicted value.
Exemplary method
An exemplary embodiment of the present disclosure provides a media object lifecycle stage prediction method, which may be applied in a server, as shown in fig. 2, and may include the following steps S201 to S205:
step S201, acquiring a characteristic data set associated with a media object to be predicted and a target user;
step S202, inputting the characteristic data set into a pre-trained stage prediction model to obtain a first prediction value of the media object to be predicted;
in the embodiment of the present disclosure, the first predicted value is used to represent an initial predicted value of a life cycle stage in which the media object to be predicted is currently located.
Step S203, determining the user weight of the target user;
in the embodiment of the present disclosure, the user weight is used to represent the importance degree of the user for predicting the life cycle stage of the media object to be predicted.
Step S204, determining a second predicted value of the media object to be predicted according to the first predicted value and the user weight;
in the embodiment of the present disclosure, the second predicted value is used to represent a final predicted value of a life cycle stage in which the media object to be predicted is currently located.
And step S205, determining the current life cycle stage of the media object to be predicted according to the second predicted value.
To sum up, the media object lifecycle stage prediction method provided by the embodiment of the present disclosure provides a user-dimensional media object lifecycle stage prediction method, which may determine an initial prediction value of a lifecycle stage in which a media object to be predicted is currently located based on a feature data set associated with the media object to be predicted and a target user, and further determine a final prediction value of the lifecycle stage in which the media object to be predicted is currently located by combining the initial prediction value and a weight of the user in a media object lifecycle stage prediction process, so as to improve accuracy of a prediction result of the media object lifecycle stage.
In an alternative embodiment, in step S201, the server may obtain a feature data set of the media object to be predicted associated with the target user.
In the embodiment of the present disclosure, the media object to be predicted is a media object that needs to be predicted in the current lifecycle stage, and the media object may be audio or video. The target user is a user selected to be used for predicting the current life cycle stage of the media object in all playing users associated with the media object to be predicted; the association feature data set of the media object to be predicted and the target user may include a media object feature data set of the media object to be predicted, a user feature data set of the target user, and a user behavior feature data set of the target user for the media object to be predicted.
In an alternative embodiment, the process of determining the target user may include: and determining the playing information of all playing users related to the media object to be predicted at the current moment, and determining whether the playing user is the target user or not according to the playing information for each playing user. The current time may be the time when the server acquires the life cycle prediction request, and the playing information may include a playing duration, a playing frequency, and/or a first playing time.
The process of determining whether the playing user is the target user according to the playing information of each playing user may include: if the playing time length is longer than a first preset time length, determining that the playing user is a target user; or, if the time difference between the first playing time and the online time of the media object to be predicted is smaller than the time threshold, and the playing time is greater than the preset playing time, it may be determined that the playing user is the target user. The first preset duration and the preset playing times may be determined based on actual needs, which is not limited in the embodiments of the present disclosure.
In an alternative embodiment, the first prediction value of the media object to be predicted needs to be determined based on a phase prediction model, and in the training process of the phase prediction model, for the sample media object, a target user of the sample media object may be determined, and the phase prediction model is trained by using a sample feature data set associated with the target user. In the actual application process of the phase prediction model, in order to improve the accuracy of the life cycle phase prediction value of the determined media object to be predicted, a target user of the sample media object may be determined as a target user of the media object to be predicted, and then the process of determining the target user may include: acquiring all playing users associated with the determined sample media object in the training process of the stage prediction model; acquiring playing information of each playing user to the sample media object at the moment of acquiring the sample characteristic data set; and determining whether the playing user is the target user according to the playing information, wherein the playing information is the target user. The playing information may include a playing time length, a playing time number and/or a first playing time, etc. For the media object to be predicted, the life cycle stage of the sample media object to be predicted can be predicted by using the feature data set associated with the target user which is the same as the training stage prediction model, and the accuracy of the determined life cycle stage of the media object to be predicted is further improved.
The process of determining whether the playing user is the target user according to the playing information may include: if the playing time length is longer than a first preset time length, determining that the playing user is a target user; or, if the time difference between the first playing time and the online time of the sample media object is smaller than the time threshold, and the playing time is greater than the preset playing time, it may be determined that the playing user is the target user.
In an alternative embodiment, the process of obtaining the feature data set of the media object to be predicted associated with the target user may include: acquiring a media object characteristic data set of a media object to be predicted, a user characteristic data set of a target user and a user behavior characteristic data set of the target user aiming at the media object to be predicted; and determining a media object characteristic data set of the media object to be predicted, a user characteristic data set of the target user and a user behavior characteristic data set of the target user aiming at the media object to be predicted as a characteristic data set associated with the media object to be predicted and the target user. For the media object to be predicted, the feature information of the media object to be predicted, the user feature information of the target user and the behavior feature information of the target user to be predicted can be combined into a multi-feature data set, and the accuracy of the life cycle stage of the determined media object to be predicted is further improved.
Wherein the media object feature data set of the media object to be predicted may comprise: the online time of the media object to be predicted, the label of the media object, the ring ratio of the average effective playing times, the interaction rate, the distribution of playing paths and the like. For audio, the media object tag may include the music, the year, the language and/or artist information of the media object, and for video, the media object tag may include the video type, the video duration and/or the video issuer information; the ring ratio of the average effective playing times may include a cycle ratio of the average effective playing times, a month-to-ring ratio of the average effective playing times and/or a month-to-cycle ratio of the average effective playing times, where the effective playing times refer to a playing behavior in which the playing duration is greater than a second preset duration, and the second preset duration may be determined based on actual needs, which is not limited in the embodiments of the present disclosure. For example, for audio, the second preset duration may be 30 seconds; the interaction rate may include a collection rate, an approval rate and a forwarding rate, where the collection rate may be a ratio of the playing times of the collection behavior to the total number of played, and the approval rate, the forwarding rate and the collection rate are determined in a similar manner, which is not described in detail in this disclosure. The playing path refers to a channel where the media object to be predicted is played by the target user, for example, the media object to be predicted is played by searching for the media object to be predicted, or the media object to be predicted is played by clicking a sharing link of another person, or the media object to be predicted is played by clicking recommendation information of a media object platform.
The user characteristic data set of the target user may include demographic attributes, content asset attributes, portrait tags, media object preferences, and/or the like of the target user, wherein the demographic attributes may include account identification, gender, age, occupation, education level, and/or constellation, and the like of the target user, the content asset attributes may include the number of content assets of the target user on the media object platform, the number of content assets may be the number of created media object lists, the number of comments posted for the media object, the number of media object uploads, and/or the number of dynamic postings, and the like. The portrait label is determined based on the playing information of the target user on the media object, for example, for the target user with more singing lists, the portrait label of the target user can be determined as the singer; for a target user who frequently uploads popular science videos, the portrait label of the target user can be determined to be a popular science person and the like; media object preferences are the degree of preference of a target user for different types or genres of media objects.
The user behavior feature data set of the target user for the media object to be predicted may include a first playing date of the target user for the media object to be predicted, a first playing path of the target user for the media object to be predicted, a last playing date of the target user for the media object to be predicted, a last playing path of the target user for the media object to be predicted, a first interaction time of the target user for the media object to be predicted, for example, a first collection time, a first sharing time, and the like of the target user for the media object to be predicted, a playing number of days, a playing number, a playing duration, and the like of the target user for the media object to be predicted.
In an alternative embodiment, the process of the server obtaining the feature data set of the media object to be predicted associated with the target user may include: determining a characteristic data set acquisition period, determining characteristic data set acquisition time in the characteristic data set acquisition period, and acquiring a characteristic data set of the media object to be predicted and the target user at the characteristic data set acquisition time. The characteristic data set acquisition period is the time between the start time of the characteristic data set acquisition period and the end time of the characteristic data set acquisition period, the end time of the characteristic data set acquisition period can be the current time or any time before the current time, and the start time of the characteristic data set acquisition period is a time which is a third preset time length earlier than the end time of the characteristic data set acquisition period. The feature data set acquisition time may be any time in the feature data set acquisition period. For example, the end time of the feature data set acquisition cycle is 2022 years, 3 months, 15 days, 10 points, 0 min and 0 sec; and if the third preset time length is 10 days, the starting moment of the feature data set acquisition period is 2022 years, 3 months, 6 days, 10 points, 0 and 0. It should be noted that the starting time of the feature data set acquisition cycle cannot be earlier than the online time of the song to be predicted, and the third preset time period may be determined based on actual needs, which is not limited in the embodiment of the present disclosure.
In an alternative embodiment, in step S202, the server may input the feature data set into a pre-trained stage prediction model to obtain a first predicted value of the media object to be predicted.
In the embodiment of the disclosure, the phase prediction model may predict the first predicted value of the life cycle phase in which the media object to be predicted is currently located based on the feature data set of the target user and the pre-trained phase prediction model, and since the determination process of the initial predicted value takes into account the feature information of the target user, the accuracy of the determined first predicted value may be improved.
In an alternative embodiment, as shown in fig. 3, the phase prediction model training method may include steps S301 to S305:
step S301, collecting a sample characteristic data set of a sample media object and a target user;
in embodiments of the present disclosure, a sample media object is a media object used to train a phase prediction model, the sample media object being a media object with a complete life cycle phase, the life cycle typically comprising a plurality of life cycle phases, e.g., a life cycle may comprise a silent phase, a growing phase and a declining phase. The sample feature dataset associated with the sample media object and the target user may include a media object feature dataset for the sample media object, a user feature dataset for the target user, and a user behavior feature dataset for the target user for the sample media object.
In an alternative embodiment, the process of collecting a sample feature data set of a sample media object associated with a target user may comprise: and in the sample characteristic data set acquisition period, determining the sample characteristic data set acquisition time, and acquiring the characteristic data set of the sample media object associated with the target user at the sample characteristic data set acquisition time.
The sample characteristic data set acquisition period may be a time between a start time of the sample characteristic data set acquisition period and an end time of the sample characteristic data set acquisition period, and the sample characteristic data set acquisition time may be any time in the sample characteristic data set acquisition period. For obtaining a richer sample feature data set, the sample feature data set acquisition period starting time may be a starting time of a life cycle of the sample media object, and the sample feature data set acquisition period ending time may be an ending time of the life cycle of the sample media object, for example, if the life cycle of the sample media object includes a silent phase, a growing phase and a decay phase, the sample feature data set acquisition period may include a time from the starting time of the silent phase of the sample media object to the ending time of the decay phase of the sample media object. The sample feature data set acquisition time may be any of the different lifecycle phases of the lifecycle of the sample media object.
For example, as shown in fig. 4, fig. 4 shows a time information diagram of different life cycle phases of a sample media object, wherein the abscissa is time, the ordinate is the average effective playing time of the sample media object in 7 days, the life cycle of the sample media object includes a silent phase 401, a growing phase 402 and a fading phase 403, wherein the start time of the silent phase 401 is 10 months and 1 days, and the end time of the silent phase 401 is 11 months and 19 days; the starting time of the growth phase 402 is 11 months and 20 days, and the ending time of the growth phase 402 is 12 months and 8 days; the beginning of the decay phase 403 is 12 months and 9 days, and the end of the decay phase 403 is 1 month and 7 days in the next year. In order to obtain a richer sample characteristic data set, the collection of the average effective playing times of 7 days in the sample characteristic data set can be carried out in the silent stage from 10 months 1 to 11 months 19 days, in the growth stage from 11 months 20 to 12 months 8 days, and in the decline stage from 12 months 9 to the next year from 1 month 7 days.
Step S302, determining a life cycle stage label value of the sample media object according to the life cycle information of the sample media object and the sample characteristic data set;
in the embodiment of the present disclosure, the playing behavior of the target user for the sample media object may generally involve multiple life cycle stages of the sample media object, and for the sample media object associated with the target user, the tag value of the sample media object cannot be determined as one of the multiple life cycle stages, and the life cycle stage tag value of the sample media object needs to be determined in combination with the life cycle information of the sample media object and the time information of the target user playing the sample media object at different life cycle stages in the sample feature data set.
In an alternative embodiment, the process of the server determining the life cycle phase tag value of the sample media object according to the life cycle information of the sample media object and the sample feature data set may include: obtaining playing time information of a target user on a sample media object in a sample characteristic data set; determining the playing time length of the sample media object in at least one target life cycle stage by the target user according to the playing time information and the starting time of each life cycle stage of the sample media object; determining the playing time ratio of the target user to the sample media object in the target life cycle stage according to the playing time of the target user in each target life cycle stage and the time of the target life cycle stage; and determining the life cycle stage label value of the sample media object according to the ratio of the coding value of each target life cycle stage to the playing time length of the target user to the sample media object in the target life cycle stage. The target life cycle stage is a life cycle stage in which a target user has a play behavior of the sample media object in a plurality of life cycle stages of the sample media object. The encoded values of the life cycle phases are identification values set for different life cycle phases in the life cycle of the media object. The encoding value of the life cycle stage can be determined based on actual needs, which is not limited by the embodiments of the present disclosure. In the process of determining the life cycle stage label value of the sample media object, the playing behavior of the sample media object by the user is taken into consideration, and the life cycle stage label value which is more in line with the actual playing condition of the sample media object can be determined.
It should be noted that, since the life cycle phases of the media object are a continuously changing process, for example, if the life cycle includes a silent phase, a growing phase and a declining phase, the life cycle of the media object generally enters the growing phase from the silent phase and then enters the declining phase from the growing phase. Therefore, in the process of determining the encoding values of the life cycle phases, the encoding values are determined for a plurality of life cycle phases in an ascending order according to the sequence of the life cycle phases that the media object can experience. For example, if the life cycle includes a silent phase, a growing phase and a declining phase, the code value for the silent phase may be 0, the code value for the growing phase may be 1 and the code value for the declining phase may be 2; alternatively, the coding value for the silent phase may be 1, the coding value for the growth phase may be 2, and the coding value for the decline phase may be 3.
In an alternative embodiment, the process of determining the life cycle stage tag value of the sample media object according to the ratio of the code value of each target life cycle stage to the playing time length of the target user to the sample media object in the target life cycle stage may include: for each target life cycle stage, determining the product of the coding value of the target life cycle stage and the playing time length ratio of the target user to the sample media object in the target life cycle stage, determining the third sum of the product corresponding to each target life cycle stage, determining the fourth sum of the playing time length ratio of each target life cycle stage, and determining the ratio of the third sum to the fourth sum to obtain the life cycle stage label value of the sample media object.
By way of example, assume that the life cycle of a sample media object includes a silent period, a growth period, and a decay period, the silent period having a code value of 1, the growth period having a code value of 2, and the decay period having a code value of 3, the silent period of the sample media object includes 10 months 1 to 11 months 19 days, the growth period includes 11 months 20 to 12 months 8 days, and the decay period includes 12 months 9 to the next year 1, 7 days. Wherein the user starts playing the sample media object on day 4 of 12 months and does not play the sample media object any more on day 27 of 12 months and 9 days. It may be determined that the target user played the sample media object during the growth and decay phases of the sample media object, the target lifecycle stages associated with the target user are the growth stage and the decay stage of the sample media object, for the sample media object, the target user's play duration ratio during the growth phase is about 1/4, the target user's play duration ratio during the decline phase is 1/2, it may be determined that the product of the encoded value 2 for the growth phase and the play-time-length ratio of the target user in the growth phase of about 1/4 is 1/2, the product of the encoded value 3 for the fade phase and the play-time-length ratio of the target user in the fade phase of 1/2 is 3/2, and the third sum value 2 of the product 1/2 corresponding to the growth phase and the product 3/2 corresponding to the fade phase is 2; a fourth sum 3/4 is determined, wherein the play time ratio of the target user in the growth phase is about 1/4 and the play time ratio in the decay phase is 1/2, and the ratio of the third sum 2 to the fourth sum 3/4 is determined, so that the life cycle phase tag value of the sample media object is about 2.67.
Step S303, inputting the sample characteristic data set into a stage prediction model to be trained to obtain a first sample prediction value of a sample media object;
in an embodiment of the present disclosure, the plurality of lifecycle phases in the lifecycle of the media object is a continuously changing process. Therefore, determining the life cycle stage of the media object is not simply to distinguish which life cycle stage of the life cycle the media object is in, but rather to determine a specific sub-stage of the life cycle stage of the media object is in, for example, an initial stage, an intermediate stage or a later stage of the life cycle the media object is in, and thus, the stage prediction model may be a linear Regression model, for example, the stage prediction model may be a light gbm (light Gradient Boosting machine) model, a Logistic Regression (LR) model or a Generalized Linear Model (GLM).
In an optional embodiment, the sample feature data set is input into a stage prediction model to be trained to obtain a sample first prediction value of the sample media object.
It should be noted that, before inputting the sample feature data set into the phase prediction model to be trained, the sample feature data set may be preprocessed, and the process of preprocessing the sample feature data set may include: the digital data in the sample characteristic data set is normalized, and the literal data in the sample characteristic data set is subjected to one-hot-Encoding (Onehot-Encoding), so that the stage prediction model can conveniently process the sample characteristic data set, and the efficiency of obtaining the first predicted value of the sample media object is improved.
Step S304, determining a loss function value according to the life cycle stage label value and the sample first predicted value;
in an embodiment of the disclosure, the loss function value is a function of the loss function value for the life cycle stage tag value of the sample media object and the sample first prediction value output by the stage prediction model, and the loss function value is used to evaluate the life cycle stage tag value of the sample media object and the degree of difference of the sample first prediction value output by the stage prediction model. Whether the model is trained can be evaluated according to the loss function value. It should be noted that in the embodiment of the present disclosure, the type of the loss function may be determined according to actual needs, and for example, the loss function may be a Mean Square Error loss function (MSE). Wherein the mean square error loss function is:
Figure BDA0003590491310000141
where n denotes the number of sample feature data sets, yiThe life cycle stage label value of the ith sample data set in the n sample characteristic data sets is represented,
Figure BDA0003590491310000142
and representing the sum sample first predicted value of the ith sample data set in the n sample characteristic data sets.
In an alternative embodiment, the process of determining the loss function value based on the life cycle stage tag value and the sample first prediction value may include: and processing the life cycle stage label value and the first sample predicted value based on the loss function to obtain a loss function value.
In step S305, the model parameters of the phase prediction model are updated according to the loss function values.
In the embodiment of the present disclosure, the sample feature data set used for training the phase prediction model generally includes a plurality of sample feature data sets, the sample feature data set is related to the media object playing behavior of the target user, the media object playing behaviors of different target users are different, and the importance degrees of the sample feature data generated by the different media object playing behaviors in the media object life cycle prediction process are different; in the stage model training process, if the sample feature data sets generated by the playing behaviors of different media objects are processed without difference, the degree of adjustment of model parameters based on each sample feature data set is consistent in the stage model training process, so that the accuracy of a first predicted value obtained by a stage prediction model obtained by training in the actual application process is low. Therefore, the sample weight can be set for each sample characteristic data set, so that the model parameters can be adjusted conveniently according to the importance degree of each sample characteristic data set in the stage model training process, and the accuracy of the first predicted value obtained by the stage prediction model in the actual application process is improved.
The importance degree of the sample characteristic data associated with the target user in the life cycle prediction process of the media object is determined due to the media object playing behavior of the target user, so that the sample weight of the sample characteristic data set associated with the target user can be determined based on the media object playing behavior information of the target user, wherein the media object playing behavior information can include the number of played media objects of the target user in the sample characteristic data set acquisition period; or the time length of the media object played by the target user in the sample characteristic data set acquisition period, and the like.
In an alternative embodiment, the media object playing behavior information includes the number of played media objects of the target user in the sample feature data set acquisition period, and the process of determining the sample weight of the sample feature data set associated with the target user based on the media object playing behavior information of the target user may include: acquiring the number of played media objects of a target user in a sample characteristic data set acquisition period; sample weights for the sample feature data sets associated with the target user are determined based on the number of media objects played. It should be noted that the media objects that have been played by the target user in the sample feature data set acquisition period include all media objects that have been played by the target user in the sample feature data set acquisition period. The sample weight can be determined based on the number of played media objects of the user, and the more credible sample weight can be determined by considering the user behavior in the determination process of the sample weight, so as to train a stage prediction model which can more accurately determine the first prediction value of the media object to be predicted.
It should be noted that, if the media object playing behavior information includes the number of played media objects in the sample feature data set acquisition period of the target user, in practical applications, when the number of played media objects in the sample feature data set acquisition period of the target user is greater or less, the importance degree of the sample feature data associated with the target user in the media object life cycle prediction process is not greatly different, so that the sample weight of the sample feature data set associated with the target user may be determined based on a sigmoid function, and the process of determining the sample weight of the sample feature data set associated with the target user according to the number of played media objects may include obtaining the sample weight of the sample feature data set associated with the target user according to the number of played media objects and the first formula. Wherein the first formula is:
weight＝1/(1+e-(songs-250)/50)；
where weight is the sample weight and songs is the number of played media objects.
As shown in fig. 5, fig. 5 is a diagram illustrating a relationship between the number of played media objects of a target user and a sample weight, wherein the sample weight is determined based on a sigmoid function and the number of played media objects. As can be seen from fig. 5, when the number of played media objects is small, the corresponding sample weight difference is small, when the number of played media objects is increased, the corresponding sample weight difference is also increased, and when the number of played media objects is large, the corresponding sample weight difference is also small, and it can be seen that the sample weights corresponding to the number of played media objects, which are determined based on the sigmoid function, conform to the variation trend of the importance degree of the sample feature data associated with the target user in the media object life cycle prediction process in practical application.
It should be further noted that, in the embodiment of the present disclosure, the model parameters include a first model parameter and a second model parameter, the first model parameter is a parameter that needs to be learned in the training process of the stage prediction model, such as a coefficient in a linear regression model, or a weight and a bias in a neural network, and the second model parameter is a parameter used to determine the stage prediction model, for example, a learning rate, an iteration number, a network layer number, a number of neurons in each layer, and the like of the stage prediction model, where the sample weight belongs to the second model parameter. It is to be understood that, after the training of the phase prediction model is started, the sample weights in the second model parameters may be updated by using the sample weights of the sample feature data set in the process of inputting the sample feature data set into the phase prediction model to be trained.
In an alternative embodiment, since the training process of the phase model considers the sample weights of the sample feature data sets, in each training process, the loss function needs to be multiplied by the sample weights of the sample feature data sets used in the iterative training process, and the process of updating the model parameters of the phase prediction model according to the loss function values may include: if the loss function value is larger than or equal to the preset loss function threshold value, determining the loss function and the sample weight to obtain an updated loss function, then determining the gradient value of the updated loss function relative to the stage learning model, then determining the product of the gradient value and the model learning rate, and finally determining the difference value between the model parameter and the product of the stage prediction model to obtain the updated model parameter of the stage prediction model. And repeating the steps S303 to S305 until the loss function value is smaller than the preset loss function threshold, and determining that the stage prediction model is trained completely.
It should be noted that, in the embodiment of the present disclosure, in the training process of the phase prediction model, in the process of determining the life cycle phase tag value of the sample book feature data set, the playing behavior of the user on the sample media object is taken into consideration, so that the life cycle phase tag value more conforming to the actual playing condition of the sample media object can be determined, and the accuracy of the first predicted value of the media object to be predicted, which is determined in the actual application of the trained phase prediction model, can be further improved.
In an alternative embodiment, the server may input the feature data set into a pre-trained phase prediction model to obtain a first predicted value of the media object to be predicted.
In an alternative embodiment, in step S203, the server may determine the user weight of the target user.
In an embodiment of the disclosure, the user weight relates to a loss function value determined based on a sample feature data set associated with the target user during model training.
In an alternative embodiment, the process of the server determining the user weight of the target user may include: obtaining a loss function value determined based on a sample feature data set associated with a target user in a training process of a stage prediction model; acquiring the number of sample media objects which are acquired in a sample characteristic data set acquisition period and are associated with a target user; and determining the user weight of the target user according to the loss function value and the number of the sample media objects. The user weight may be determined by a loss function value associated with the target user, resulting in a user weight that more closely conforms to the actual degree of influence of the target user.
Wherein the sample feature data set associated with the target user includes a plurality of values, determining the user weight of the target user based on the loss function value and the number of sample media objects may include: determining a total value of the loss function values based on the sum of the loss function values corresponding to the target user; then determining the ratio of the total value of the loss function value to the number of the sample media objects; and finally, determining the reciprocal of the ratio to obtain the user weight of the target user. The user weight may be determined by a plurality of loss function values associated with the target user, resulting in a user weight that more closely conforms to the actual degree of influence of the target user.
For example, the process of determining the user weight of the target user based on the loss function value and the number of sample media objects may include: and obtaining the user weight of the target user according to the loss function value, the number of the sample media objects and the second formula. Wherein the second formula is:
Figure BDA0003590491310000181
wherein, W is the user weight of the target user, Li is m sample characteristic data sets associated with the target user, the loss function value corresponding to the ith sample characteristic data set, and S is the number of sample media objects associated with the target user and acquired in the sample characteristic data set acquisition period.
In an alternative embodiment, after the phase prediction model is trained, the user weight of each target user may be predetermined based on the above manner, and the user weight information table is generated, and the process that the server may determine the user weight of the target user may include: and inquiring and acquiring the user weight of the target user in the user weight information table.
In an alternative embodiment, in step S204, the server may determine the second predicted value of the media object to be predicted according to the first predicted value and the user weight.
In the embodiment of the present disclosure, for the media object to be predicted, the actual influence degree of the target user on the determination of the current life cycle stage of the media object to be predicted needs to be considered, so as to improve the accuracy of the second predicted value of the determined media object to be predicted.
In an optional implementation manner, the number of the target users is at least two, and the determining a second predicted value of the media object to be predicted according to the predicted value of the lifecycle stage to be processed and the user weight includes: determining a first sum of the products of the first predicted values and the user weights associated with each target user; determining a second sum of user weights for a plurality of target users; and determining the ratio of the first sum value to the second sum value to obtain a second predicted value of the media object to be predicted. The second predicted value of the media object to be predicted can be determined by utilizing the feature data sets and the user weights of the target users, and the accuracy of the determined final predicted value of the life cycle stage where the media object to be predicted is located at present is further improved.
For example, the process of determining the second predicted value of the media object to be predicted according to the predicted value of the lifecycle stage to be processed and the user weight may include: determining a second predicted value of the media object to be predicted according to the predicted value of the life cycle stage to be processed, the user weight and a third formula, wherein the third formula is as follows:
Figure BDA0003590491310000191
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003590491310000192
is a second predicted value, P, of the media object to be predictediIs the first predicted value, W, of the ith target user in the plurality of target users tiThe user weight of the ith target user in the plurality of target users t.
In an alternative embodiment, in step S205, the server may determine the current life cycle stage of the media object to be predicted according to the second predicted value.
In this embodiment of the present disclosure, the life cycle of the media object may include multiple life cycle phases, and the current life cycle phase of the media object to be predicted may be determined according to the second predicted value of the media object to be predicted and the encoded values corresponding to the multiple life cycle phases of the media object, respectively.
In an alternative embodiment, the process of the server determining the current life cycle stage of the media object to be predicted according to the second prediction value may include: acquiring the encoding value of each life cycle stage of the media object, determining at least one encoding value which is less than or equal to the second predicted value in the encoding values of each life cycle stage of the media object to obtain at least one first candidate encoding value, determining the maximum encoding value in the at least one first candidate encoding value to obtain a first target encoding value, and determining the life cycle stage corresponding to the first target encoding value as the current initial life cycle stage of the media object to be predicted.
Further, the encoding value of at least one life cycle sub-stage of the target life cycle stage is obtained, at least one encoding value which is smaller than or equal to the second predicted value in the encoding values of the at least one life cycle sub-stage is determined, at least one second candidate encoding value is obtained, the maximum encoding value in the at least one second candidate encoding value is determined, the second target encoding value is obtained, and the life cycle sub-stage corresponding to the second target encoding value is determined as the current life cycle stage of the media object to be predicted. The target life cycle stage is a life cycle stage corresponding to the first target code value.
For example, if the life cycle stages of the media object include a silent stage, a growing stage and a declining stage, the code value of the silent stage is 1, the code value of the growing stage is 2, the code value of the declining stage is 3, and the second predicted value is 2.89, it may be determined that the code value 1 of the silent stage and the code value 2 of the growing stage are less than or equal to the second predicted value 2.89, the code value 1 of the silent stage and the code value 2 of the growing stage are determined as two first candidate code values, and since the code value 2 of the growing stage is the largest code value of the two first candidate code values, it may be determined that the song to be predicted is currently in the growing stage.
Further, it can be determined that the target lifecycle stage is a growth stage, wherein the code values of the three lifecycle sub-stages of the growth stage are respectively 2.00 for the initial stage of the growth stage, 2.33 for the middle stage of the growth stage, and 2.66 for the later stage of the growth stage. Since the encoding value 2.00 of the initial stage of the growing stage, the encoding value 2.33 of the middle stage of the growing stage and the encoding value 2.66 of the later stage of the growing stage are all less than or equal to the second predicted value 2.89, the encoding value 2.00 of the initial stage of the growing stage, the encoding value 2.33 of the middle stage of the growing stage and the encoding value 2.66 of the later stage of the growing stage are determined as three second candidate encoding values, and since the encoding value 2.66 of the later stage of the growing stage is the largest encoding value of the three second candidate encoding values, it can be determined that the song to be predicted is currently in the later stage of the growing stage.
Exemplary devices
Having described the method of the exemplary embodiment of the present disclosure, the apparatus of the exemplary embodiment of the present disclosure will be described next with reference to fig. 6 and 7.
An embodiment of the present disclosure provides a media object lifecycle stage prediction apparatus, as shown in fig. 6, a lifecycle stage prediction apparatus 600 includes:
a first obtaining module 601 configured to obtain a feature data set of a media object to be predicted associated with a target user;
a first prediction module 602, configured to input the feature data set into a pre-trained stage prediction model to obtain a first prediction value of a media object to be predicted;
a first determining module 603 configured to determine a user weight of the target user;
a second determining module 604 configured to determine a second predicted value of the media object to be predicted according to the first predicted value and the user weight;
a third determining module 605 configured to determine the current life cycle stage of the media object to be predicted according to the second predicted value.
Optionally, the number of target users is at least two, and the second determining module 604 is configured to:
determining a first sum of the products of the first predicted values and the user weights associated with each target user;
determining a second sum of user weights for a plurality of target users;
and determining the ratio of the first sum value to the second sum value to obtain a second predicted value of the media object to be predicted.
Optionally, the first determining module 603 is configured to:
obtaining a loss function value determined based on a sample feature data set associated with a target user in a training process of a stage prediction model;
acquiring the number of sample media objects which are acquired in a sample characteristic data set acquisition period and are associated with a target user;
and determining the user weight of the target user according to the loss function value and the number of the sample media objects.
Optionally, the number of sample feature data sets associated with the target user includes a plurality, and the first determining module 602 is configured to:
determining a total value of the loss function values based on the sum of the loss function values corresponding to the target user;
determining a ratio of the total value of the loss function values to the number of sample media objects;
and determining the reciprocal of the ratio to obtain the user weight of the target user.
Optionally, the first obtaining module 601 is configured to:
acquiring a media object characteristic data set of a media object to be predicted, a user characteristic data set of a target user and a user behavior characteristic data set of the target user aiming at the media object to be predicted;
and determining the media object characteristic data set, the user characteristic data set and the user behavior characteristic data set as the characteristic data set of the media object to be predicted associated with the target user.
Optionally, as shown in fig. 6, the life cycle stage predicting apparatus 600 further includes. A fourth determination module 606 configured to:
acquiring all playing users associated with the determined sample media object in the training process of the stage prediction model;
acquiring the playing information of each playing user on the sample media object;
and determining whether the playing user is the target user according to the playing information.
An embodiment of the present disclosure provides a phase prediction model training apparatus, as shown in fig. 7, a phase prediction model training apparatus 700 includes:
an acquisition module 701 configured to acquire a sample feature dataset of a sample media object associated with a target user;
a fifth determining module 702 configured to determine a life cycle phase tag value of the sample media object based on the life cycle information of the sample media object and the sample feature data set;
the model training module 703 is configured to input the sample feature data set into a stage prediction model to be trained, so as to obtain a sample first prediction value of the sample media object;
a sixth determining module 704 configured to determine a loss function value based on the life cycle stage tag value and the sample first predicted value;
a model update module 705 configured to update model parameters of the phase prediction model according to the loss function values.
Optionally, the lifecycle includes a plurality of lifecycle stages, the lifecycle information includes a start time and a duration of each lifecycle stage, and the fifth determining module 702 is configured to:
obtaining playing time information of a target user on a sample media object in a sample characteristic data set;
determining the playing time length of the sample media object in at least one target life cycle stage by the target user according to the playing time information and the starting time of each life cycle stage of the sample media object;
determining the playing time ratio of the target user to the sample media object in the target life cycle stage according to the playing time of the target user in each target life cycle stage and the time of the target life cycle stage;
and determining the life cycle stage label value of the sample media object according to the ratio of the coding value of each target life cycle stage to the playing time length of the target user to the sample media object in the target life cycle stage.
Optionally, the model parameters include sample weights, and as shown in fig. 7, the phase model training apparatus further includes a seventh determining module 706 configured to:
acquiring the number of played media objects of a target user in a sample characteristic data set acquisition period;
the sample weight is determined based on the number of media objects that have been played.
In addition, other specific details of the embodiments of the present disclosure have been described in detail in the embodiments of the invention of the above method, and are not described herein again.
Exemplary storage Medium
The storage medium of the exemplary embodiment of the present disclosure is explained below.
In the present exemplary embodiment, the above-described method may be implemented by a program product, such as a portable compact disc read only memory (CD-ROM) and including program code, and may be executed on a device, such as a personal computer. However, the program product of the present disclosure is not limited thereto, and in this document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
A computer readable signal medium may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RE, etc., or any suitable combination of the foregoing.
Program code for carrying out operations for the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a local area network (FAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., through the internet using an internet service provider).
Exemplary electronic device
An electronic device of an exemplary embodiment of the present disclosure is explained with reference to fig. 8.
The electronic device 800 shown in fig. 8 is only an example and should not bring any limitations to the functionality and scope of use of the embodiments of the present disclosure.
As shown in fig. 8, electronic device 800 is in the form of a general purpose computing device. The components of the electronic device 800 may include, but are not limited to: at least one processing unit 810, at least one memory unit 820, a bus 830 connecting the various system components (including the memory unit 820 and the processing unit 810), and a display unit 840.
Where the memory unit stores program code, the program code may be executed by the processing unit 810 to cause the processing unit 810 to perform steps according to various exemplary embodiments of the present disclosure as described in the "exemplary methods" section above in this specification. For example, processing unit 810 may perform method steps, etc., as shown in fig. 2.
The storage unit 820 may include volatile storage units such as a random access storage unit (RAM)821 and/or a cache storage unit 822, and may further include a read only storage unit (ROM) 823.
Storage unit 820 may also include a program/utility 824 having a set (at least one) of program modules 825, such program modules 825 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.
Bus 830 may include a data bus, an address bus, and a control bus.
The electronic device 800 may also communicate with one or more external devices 900 (e.g., keyboard, pointing device, bluetooth device, etc.), which may be through an input/output (I/O) interface 850. The electronic device 800 further comprises a display unit 840 connected to the input/output (I/O) interface 850 for displaying. Also, the electronic device 800 may 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 the network adapter 860. As shown, the network adapter 860 communicates with the other modules of the electronic device 800 via the bus 830. It should be appreciated that although not shown, other hardware and/or software modules may be used in conjunction with the electronic device 800, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.
It should be noted that although in the above detailed description several modules or sub-modules of the apparatus are mentioned, such 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, in accordance with embodiments of the present disclosure. Conversely, the features and functions of one unit/module described above may be further divided into embodiments by a plurality of units/modules.
Further, while the operations of the disclosed methods are depicted in the drawings in a particular order, this does not require or imply that these operations must be performed in this particular order, or that all of the illustrated operations must be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step execution, and/or one step broken down into multiple step executions.
While the spirit and principles of the present disclosure have been described with reference to several particular embodiments, it is to be understood that the present disclosure is not limited to the particular embodiments disclosed, nor is the division of aspects, which is for convenience only as the features in such aspects may not be combined to benefit. The disclosure is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims (10)

1. A method for predicting a life cycle phase of a media object, comprising:
acquiring a characteristic data set associated with a media object to be predicted and a target user;
inputting the characteristic data set into a pre-trained stage prediction model to obtain a first prediction value of the media object to be predicted;
determining a user weight of the target user;
determining a second predicted value of the media object to be predicted according to the first predicted value and the user weight;
and determining the current life cycle stage of the media object to be predicted according to the second predicted value.
2. The method of claim 1, wherein the number of target users is at least two, and wherein determining a second predicted value of the media object to be predicted according to the to-be-processed lifecycle stage predicted value and the user weight comprises:
determining a first sum of the products of the first predicted values and the user weights associated with each of the target users;
determining a second sum of user weights for a plurality of the target users;
and determining the ratio of the first sum value to the second sum value to obtain a second predicted value of the media object to be predicted.
3. The method of claim 1, wherein the user weight of the target user is determined by:
obtaining a loss function value determined based on a sample feature data set associated with the target user in a training process of the stage prediction model;
acquiring the number of sample media objects which are acquired in a sample characteristic data set acquisition period and are associated with the target user;
and determining the user weight of the target user according to the loss function value and the number of the sample media objects.
4. A phase prediction model training method is characterized by comprising the following steps:
collecting a sample characteristic data set of a sample media object associated with a target user;
determining a life cycle stage tag value of the sample media object according to the life cycle information of the sample media object and the sample characteristic data set;
inputting the sample characteristic data set into a stage prediction model to be trained to obtain a sample first prediction value of the sample media object;
determining a loss function value according to the life cycle stage label value and the first sample predicted value;
and updating the model parameters of the stage prediction model according to the loss function values.
5. The method of claim 4, wherein the lifecycle comprises a plurality of lifecycle stages, wherein the lifecycle information comprises a start time and a duration for each of the lifecycle stages, and wherein determining the lifecycle stage tag value for the sample media object from the lifecycle information for the sample media object and the sample feature dataset comprises:
obtaining the playing time information of the target user to the sample media object in the sample characteristic data set;
determining the playing time length of the target user to the sample media object in at least one target life cycle stage according to the playing time information and the starting time of each life cycle stage of the sample media object;
determining the playing time ratio of the target user to the sample media object in the target life cycle stage according to the playing time of the target user in each target life cycle stage and the time of the target life cycle stage;
and determining the life cycle stage label value of the sample media object according to the code value of each target life cycle stage and the play time ratio of the target user to the sample media object in the target life cycle stage.
6. The method of claim 4, wherein the model parameters include sample weights, the method further comprising:
acquiring the number of played media objects of the target user in a sample characteristic data set acquisition period;
determining the sample weight according to the number of played media objects.
7. A life cycle phase prediction apparatus, the apparatus comprising:
a first obtaining module configured to obtain a feature data set of a media object to be predicted associated with a target user;
the first prediction module is configured to input the feature data set into a pre-trained stage prediction model to obtain a first prediction value of the media object to be predicted;
a first determination module configured to determine a user weight of the target user;
a second determination module configured to determine a second predicted value of the media object to be predicted according to the first predicted value and the user weight;
and the third determination module is configured to determine the current life cycle stage of the media object to be predicted according to the second predicted value.
8. A phase prediction model training apparatus, comprising:
a collection module configured to collect a sample feature dataset of a sample media object associated with a target user;
a fifth determining module configured to determine a lifecycle stage tag value for the sample media object as a function of the lifecycle information for the sample media object and the sample feature dataset;
the model training module is configured to input the sample characteristic data set into a stage prediction model to be trained to obtain a sample first prediction value of the sample media object;
a sixth determining module configured to determine a loss function value from the lifecycle stage tag value and a sample first prediction value;
a model update module configured to update model parameters of the phase prediction model according to the loss function values.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method of one of claims 1 to 3, or of any one of claims 4 to 6.
10. An electronic device, comprising:
a processor; and
a memory for storing executable instructions of the processor;
wherein the processor is configured to perform the method of any of claims 1 to 3, or claims 4 to 6, via execution of the executable instructions.
CN202210376202.0A 2022-04-11 2022-04-11 Life cycle stage prediction and model training method, device, medium and equipment Pending CN114580790A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210376202.0A CN114580790A (en) 2022-04-11 2022-04-11 Life cycle stage prediction and model training method, device, medium and equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210376202.0A CN114580790A (en) 2022-04-11 2022-04-11 Life cycle stage prediction and model training method, device, medium and equipment

Publications (1)

Publication Number Publication Date
CN114580790A true CN114580790A (en) 2022-06-03

Family

ID=81779348

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210376202.0A Pending CN114580790A (en) 2022-04-11 2022-04-11 Life cycle stage prediction and model training method, device, medium and equipment

Country Status (1)

Country Link
CN (1) CN114580790A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117710066A (en) * 2024-02-05 2024-03-15 厦门傲凡科技股份有限公司 Financial customer recommendation method and system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117710066A (en) * 2024-02-05 2024-03-15 厦门傲凡科技股份有限公司 Financial customer recommendation method and system
CN117710066B (en) * 2024-02-05 2024-05-10 厦门傲凡科技股份有限公司 Financial customer recommendation method and system

Similar Documents

Publication Publication Date Title
CN110582025B (en) Method and apparatus for processing video
US11720804B2 (en) Data-driven automatic code review
CN110781321B (en) Multimedia content recommendation method and device
JP6785904B2 (en) Information push method and equipment
US11615485B2 (en) System and method for predicting engagement on social media
US20220408131A1 (en) Image analysis system
US20210286839A1 (en) File synchronization system
US11811708B2 (en) Systems and methods for generating dynamic conversational responses using cluster-level collaborative filtering matrices
CN112925911B (en) Complaint classification method based on multi-modal data and related equipment thereof
CN111767431A (en) Method and device for video dubbing
CN111753126B (en) Method and device for video dubbing
CN104661093A (en) Method and system for determining updates for a video tutorial
CN112182281B (en) Audio recommendation method, device and storage medium
CN113672758B (en) Song list generation method, device, medium and computing equipment
CN115618024A (en) Multimedia recommendation method and device and electronic equipment
CN113344647B (en) Information recommendation method and device
CN114580790A (en) Life cycle stage prediction and model training method, device, medium and equipment
US20240046922A1 (en) Systems and methods for dynamically updating machine learning models that provide conversational responses
CN108509442B (en) Search method and apparatus, server, and computer-readable storage medium
CN115129922A (en) Search term generation method, model training method, medium, device and equipment
US20220171985A1 (en) Item recommendation with application to automated artificial intelligence
US20230267277A1 (en) Systems and methods for using document activity logs to train machine-learned models for determining document relevance
CN111797273B (en) Method and device for adjusting parameters
US20220309054A1 (en) Dynamic updating of digital data
CN116226764A (en) Data prediction method and device, storage medium and electronic equipment

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