CN114519122A - Music recommendation method based on vehicle driving scene - Google Patents
Music recommendation method based on vehicle driving scene Download PDFInfo
- Publication number
- CN114519122A CN114519122A CN202011298551.2A CN202011298551A CN114519122A CN 114519122 A CN114519122 A CN 114519122A CN 202011298551 A CN202011298551 A CN 202011298551A CN 114519122 A CN114519122 A CN 114519122A
- Authority
- CN
- China
- Prior art keywords
- user
- song
- information
- driving
- music
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
- G06F16/63—Querying
- G06F16/635—Filtering based on additional data, e.g. user or group profiles
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
Abstract
The invention provides a music recommendation method based on a vehicle driving scene, which comprises the steps of determining personality trait information of a user to be recommended; pre-classifying the driving scenes of the vehicle according to the driving information; determining a current driving scene of the vehicle; acquiring music preference information of a user to be recommended; determining the preference degree of the user to be recommended to each song according to the music preference information of the music preference similar user, the personality trait information of the personality trait similar user, the preference information of the music preference similar user to each song, and the preference information of the personality trait similar user to each song; and constructing a music recommendation list according to the preference of the user to be recommended to each song. According to the scheme, the influence of the driving scene on the mood of the user is fully considered when the driving scene of the vehicle is determined. And the vehicle driving scenes are pre-classified according to the driving information, so that the calculation amount can be effectively reduced, and the recommendation efficiency is improved.
Description
Technical Field
The invention relates to the technical field of vehicle-mounted music recommendation methods, in particular to a music recommendation method based on a vehicle driving scene.
Background
With the wide application of big data analysis technology, the function of recommending related content based on user interest is gradually merged into the daily life of consumers. Among them, music recommendation functions and related products are getting more and more noticed and appreciated by consumers.
Most of the existing music recommendation functions analyze which music tracks the user is interested in according to the historical listening records and the user figures of the user, so that personalized music recommendation service is realized, and the product experience satisfaction of the user is improved. At present, each big music product platform has personalized music recommendation service, and recommendation algorithms of a background are different. The method is roughly classified into a music recommendation method based on music content analysis, such as dried shrimp music, and Song Taste music website. And a music recommendation method based on user community behavior analysis, such as internet music. Music recommendation methods based on user portraits and historical listening lists, such as cool dog music, QQ music. Music recommendation methods based on data such as music comment context analysis and user preference analysis, such as broad-bean radio. Although all the big internet products actively deploy resources on music recommendation products, establish music social circles and leverage huge potential commercial profits behind the music social circles, the current products are applied to mobile terminals of clients such as mobile phones and tablet computers, so that the real-time scene information of a user is not considered when the music is recommended, and the user needs to manually input the scene information even if the scene information is considered.
In fact, the scene where the user is located determines the mood of the user at the current moment, and the music that the user wants to hear is different according to different moods. In the prior art, the scene information of the user in real time is not considered when the music is recommended, so that the accuracy of the recommendation result is not high. Even if the accuracy is improved by manually inputting scene information, the experience of the user is reduced by manual operation.
Disclosure of Invention
The invention aims to solve the problem that in the prior art, the accuracy of a recommendation result is low because the real-time scene information of a user is not considered when the music is recommended. Even if the accuracy is improved by manually inputting the scene information, the experience of the user is reduced by manual operation.
In order to solve the above problems, an embodiment of the present invention discloses a music recommendation method based on a vehicle driving scene, including the following steps:
s1: determining personality trait information of a user to be recommended according to the personality evaluation tool;
s2: collecting buried point data of a vehicle, and randomly dividing the buried point data into a training set and a test set; the buried point data comprises driving information and scene information;
s3: pre-classifying the driving information of the data of the training set according to the driving information of the buried point data;
s4: determining the current driving scene of the user to be recommended according to the pre-classification result and the current driving information of the user to be recommended;
s5: carrying out correlation analysis on scene information corresponding to the current driving scene of the user to be recommended so as to obtain music preference information of the user to be recommended;
s6: determining a plurality of music preference similar users similar to the music preference of the user to be recommended according to the music preference information, and determining a plurality of personality trait similar users similar to the personality trait of the user to be recommended according to the personality trait information;
s7: determining the preference degree of the user to be recommended to each song according to the music preference information of the music preference similar user, the personality trait information of the personality trait similar user, the preference information of the music preference similar user to each song, and the preference information of the personality trait similar user to each song; and constructing a music recommendation list according to the preference of the user to be recommended to each song.
By adopting the scheme, the driving scene of the vehicle is determined according to the driving information and the scene, and the influence of the driving scene on the mood of the user is fully considered. The method has the advantages that the vehicle driving scenes are pre-classified according to the driving information, when the music recommendation list is constructed subsequently, only the type of the user to be recommended is evaluated according to the pre-classification result, and then the music recommendation list is constructed in the type according to the preference degree and the personality traits, so that the calculation amount is effectively reduced, and the recommendation efficiency is improved.
According to another specific embodiment of the invention, the music recommendation method based on the vehicle driving scene is disclosed in the embodiment of the invention, wherein
In step S1, the personality assessment tool comprises a sunny music personality assessment scale; and is
In step S2, the driving information of the buried point data includes driving speed and driving behavior habits, and the scene information of the buried point data includes weather conditions, road conditions, holiday information and song listening lists; the driving information is a continuous variable, and the scene information is a discrete variable.
By adopting the scheme, the mood of the user in the current driving scene can be comprehensively judged according to the driving speed, the driving behavior habit, the weather condition, the road condition, the holiday information, the song listening list and other factors. Therefore, when a music recommendation list is subsequently constructed, music preference of a user influenced by mood is fully considered, and a recommendation result is more accurate.
According to another specific embodiment of the present invention, in the music recommendation method based on a driving scene of a vehicle, the step S2 of collecting the buried point data of the vehicle includes:
obtaining the average running speed of the vehicle in a preset time period, and taking the average running speed as the running speed;
obtaining the hundred-kilometer driving condition of the vehicle before a preset time period, and calculating the driving behavior habit according to the following formula:
wherein S is a driving behavior habit; a. the1The number of rapid acceleration times within hundred kilometers; a. the2The number of rapid deceleration times within hundred kilometers; w is the number of sharp turns in hundred kilometers; l is1Over speed of 10% and less within hundred kilometers; l is2Over 10% and under 20% times within hundred kilometers; l is a radical of an alcohol3Over-speed of 20% and more within hundred kilometers; t is the number of times of turning to the non-turn light in hundred kilometers.
By adopting the scheme, the driving behavior habit is determined according to the times of rapid acceleration in hundred kilometers, the times of rapid deceleration in hundred kilometers, the times of rapid turning in hundred kilometers and the like, and the driving behavior habit of the user can be accurately judged. The accuracy of the recommendation result is improved.
According to another specific embodiment of the present invention, in the music recommendation method based on a driving scene of a vehicle, step S3 includes:
s31: establishing a Gaussian mixture model according to driving information of buried point data;
s32: determining a maximized model parameter of the Gaussian mixture model by using a maximum expectation algorithm;
s33: and pre-classifying the driving information of the data of the training set by utilizing Bayesian information standard.
According to another specific embodiment of the present invention, in the music recommendation method based on vehicle driving scenes disclosed in the embodiments of the present invention, in step S31, when the gaussian mixture model is established according to the driving information of the buried point data, the driving information is used as a feature column, wherein the feature column of each user to be recommended in each type of driving scene includes scene information;
step S32 includes:
s321: taking the driving speed and the driving behavior habit as sample data of a Gaussian mixture model;
s322: setting initialization model parameters for the Gaussian mixture model and repeatedly executing S323 and S324;
s323: calculating the posterior probability of the hidden variable in the Gaussian mixture model according to the initialized model parameter or the last iteration model parameter;
s324: calculating a likelihood function of the Gaussian mixture model according to the posterior probability, and maximizing the likelihood function to obtain new model parameters;
s325: stopping iteration when the difference value between the new model parameter and the iterative model parameter is smaller than a preset parameter threshold value, and taking the new model parameter when the iteration is stopped as a maximized model parameter for maximizing the likelihood function;
in step S33, the driving scenes of the vehicle are pre-classified according to the following formula:
BIC＝kln(m)-2ln(L)
wherein, BIC is the number of vehicle driving scenes; k is the number of the maximized model parameters; m is the number of samples; l is a likelihood function.
According to another specific embodiment of the invention, in the music recommendation method based on the vehicle driving scene disclosed by the embodiment of the invention, the type of the driving scene corresponding to the maximum value of the posterior probability of the hidden variable in the Gaussian mixture model is the current driving scene of the user to be recommended.
According to another specific embodiment of the present invention, after step S3, the music recommendation method based on a driving scene of a vehicle according to the embodiment of the present invention further includes:
s3': judging whether the number of songs of a song listening list of a user to be recommended in the current driving scene is smaller than a preset song number threshold value or not in the data of the training set;
if yes, determining a supplementary song with higher similarity to the song in the song listening list from the song library according to the song in the song listening list, and adding the supplementary song to the song listening list;
if not, step S4 is executed.
According to another specific embodiment of the present invention, in the method for recommending music based on a driving scene of a vehicle according to an embodiment of the present invention, in step S3', a supplemental song having a higher similarity to a song in a song listening list is determined from a song library according to the song in the song listening list, including:
determining a driving scene corresponding to the song listening list, counting the listening situations of songs in the song library of each user in the driving scene, and establishing a total listening list for the song libraries of all the users in the driving scene;
calculating the similarity coefficient of each song in the song listening list of the user to be recommended and other songs which do not exist in the song listening list in the total listening list according to the total listening list;
arranging other songs which do not exist in the song listening list in the total listening list according to the sequence of the similarity coefficient from large to small to form a supplementary song list;
and calculating the difference value between the preset song quantity threshold value and the quantity of the songs in the song listening list, and selecting the supplementary songs from the supplementary song list according to the difference value.
According to another specific embodiment of the present invention, in the music recommendation method based on a driving scene of a vehicle, step S5 includes:
s51: performing angle cosine correlation analysis on scene information corresponding to the current driving scene of the user to be recommended to acquire music preference information of all users in the current driving scene;
s52: calculating music preference similarity of all users in the current driving scene according to the music preference information;
step S6 includes:
s61: determining a plurality of music preference similar users similar to the music preference of the user to be recommended according to the nearest selection strategy; the number of users with similar music preferences is determined by a cross validation method;
s62: performing included angle cosine correlation analysis on personality trait information corresponding to the current driving scene of the vehicle to acquire personality trait information of all users in the current driving scene;
s63: calculating the personality trait similarity of all users in the current driving scene according to the personality trait information of all users in the current driving scene;
s64: determining a plurality of personality trait similar users similar to the personality traits of the user to be recommended according to the nearest selection strategy; wherein the number of users with similar personality traits is determined by a cross-validation method.
According to another specific embodiment of the present invention, in the music recommendation method based on the driving scene of the vehicle disclosed in the embodiment of the present invention, in step S7, the preference degree of the user to be recommended for each song is determined according to the following formula:
wherein, PuiThe preference degree of the user to be recommended to each song is set; sim (u, v) is the music preference similarity of user u and user v; psim (u, t) is the music personality similarity of the user u and the user t; v is a music preference similar user, and t is a personality trait similar user.
By adopting the scheme, the music preference similarity, the personality trait similarity and the preferences of the users close to the personality traits of the users to be recommended are combined by the corresponding weights, so that the accuracy of the recommendation result can be improved.
The invention has the beneficial effects that:
according to the scheme, the driving scene of the vehicle is determined according to the driving information and the scene, and the influence of the driving scene on the mood of the user is fully considered. The method has the advantages that the vehicle driving scenes are pre-classified according to the driving information, when the music recommendation list is constructed subsequently, only the type of the user to be recommended is evaluated according to the pre-classification result, and then the music recommendation list is constructed in the type according to the preference degree and the personality traits, so that the calculation amount is effectively reduced, and the recommendation efficiency is improved.
Further, in the pre-classified driving scenes, scene information such as weather conditions, road conditions, holiday information, and song listening lists, and driving information such as driving speeds and driving behavior habits are considered when calculating the music preference similarity, so that the mood of the user can be sufficiently considered when calculating the music preference.
Furthermore, the music preference similarity, the personality trait similarity and the preferences of the users close to the personality traits of the users to be recommended are combined with corresponding weights, the influence of the mood of the users on the music preference in a real-time driving scene is considered when the music preference similarity is calculated, meanwhile, the personality trait information is considered, and the recommendation accuracy is improved.
Drawings
FIG. 1 is a flowchart illustrating a music recommendation method for a driving scene of a vehicle according to an embodiment of the present invention;
FIG. 2 is another schematic flow chart of a music recommendation method for a driving scene of a vehicle according to an embodiment of the present invention;
FIG. 3 is another schematic flow chart of a music recommendation method for a driving scene of a vehicle according to an embodiment of the present invention;
FIG. 4 is another schematic flow chart of a music recommendation method for a driving scene of a vehicle according to an embodiment of the present invention;
FIG. 5 is another schematic flow chart of a music recommendation method for a driving scene of a vehicle according to an embodiment of the present invention;
fig. 6 is a table of driving information and scene information in a music recommendation method for a vehicle driving scene according to an embodiment of the present invention;
fig. 7 is a schematic diagram of a sunny music sub-personality evaluation scale in the music recommendation method for a vehicle driving scene according to the embodiment of the invention;
FIG. 8 is a schematic diagram illustrating results of pre-classification of a vehicle driving scene in the music recommendation method for a vehicle driving scene according to an embodiment of the present invention;
FIG. 9 is another schematic diagram illustrating the result of pre-classifying the driving scenes of the vehicle in the music recommendation method for the driving scenes of the vehicle according to the embodiment of the invention;
FIG. 10 is a list of listening songs of a user in a music recommendation method for a driving scene of a vehicle according to an embodiment of the present invention;
fig. 11 is a music recommendation list constructed in the music recommendation method for a vehicle driving scene according to the embodiment of the invention.
Detailed Description
The following description of the embodiments of the present invention is provided for illustrative purposes, and other advantages and effects of the present invention will become apparent to those skilled in the art from the present disclosure. While the invention will be described in conjunction with the preferred embodiments, it is not intended that features of the invention be limited to these embodiments. On the contrary, the invention is described in connection with the embodiments for the purpose of covering alternatives or modifications that may be extended based on the claims of the present invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. The invention may be practiced without these particulars. Moreover, some of the specific details have been left out of the description in order to avoid obscuring or obscuring the focus of the present invention. It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
It should be noted that in this specification, like reference numerals and letters refer to like items in the following drawings, and thus, once an item is defined in one drawing, it need not be further defined and explained in subsequent drawings.
In the description of the present embodiment, it should be noted that the terms "upper", "lower", "inner", "bottom", and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings or orientations or positional relationships that are conventionally placed when the products of the present invention are used, and are only used for convenience in describing the present invention and simplifying the description, but do not indicate or imply that the devices or elements indicated must have specific orientations, be configured in specific orientations, and operate, and thus, should not be construed as limiting the present invention.
The terms "first," "second," and the like are used solely to distinguish one from another and are not to be construed as indicating or implying relative importance.
In the description of the present embodiment, it should be further noted that, unless explicitly stated or limited otherwise, the terms "disposed," "connected," and "connected" are to be interpreted broadly, e.g., as a fixed connection, a detachable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present embodiment can be understood in specific cases by those of ordinary skill in the art.
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
The method and the device aim to solve the problem that in the prior art, the accuracy of a recommendation result is low because the real-time scene information of a user is not considered when the music is recommended. Even if the accuracy is improved by manually inputting the scene information, the experience of the user is reduced by manual operation, and the embodiment of the invention provides a music recommendation method based on a vehicle driving scene. Specifically, referring to fig. 1, the music recommendation method based on a vehicle driving scenario provided by the embodiment of the present invention specifically includes the following steps:
s1: determining personality trait information of a user to be recommended according to the personality evaluation tool;
s2: collecting buried point data of a vehicle, and randomly dividing the buried point data into a training set and a test set; the buried point data comprises driving information and scene information;
s3: pre-classifying the driving information of the data of the training set according to the driving information of the buried point data;
s4: determining the current driving scene of the user to be recommended according to the pre-classification result and the current driving information of the user to be recommended;
s5: carrying out correlation analysis on scene information corresponding to the current driving scene of the user to be recommended so as to obtain music preference information of the user to be recommended;
s6: determining a plurality of music preference similar users similar to the music preference of the user to be recommended according to the music preference information, and determining a plurality of personality trait similar users similar to the personality trait of the user to be recommended according to the personality trait information;
s7: determining the preference degree of the user to be recommended to each song according to the music preference information of the music preference similar user, the personality trait information of the personality trait similar user, the preference information of the music preference similar user to each song, and the preference information of the personality trait similar user to each song; and constructing a music recommendation list according to the preference of the user to be recommended to each song.
By adopting the scheme, the driving scene of the vehicle is determined according to the driving information and the scene, and the influence of the driving scene on the mood of the user is fully considered. The method has the advantages that the vehicle driving scenes are pre-classified according to the driving information, when the music recommendation list is constructed subsequently, only the type of the user to be recommended is evaluated according to the pre-classification result, and then the music recommendation list is constructed in the type according to the preference degree and the personality traits, so that the calculation amount is effectively reduced, and the recommendation efficiency is improved.
The following describes a music recommendation method based on a vehicle driving scenario according to an embodiment of the present invention with reference to fig. 1 to 11.
Firstly, step S1 is executed to determine personality trait information of the user to be recommended according to the personality evaluation tool.
Specifically, in this embodiment, the personality evaluation tool includes a sunny music personality evaluation scale, which may specifically refer to fig. 7.
More specifically, the sunny music sub-personality assessment scale was proposed by yoguangyu et al. The table has a total of 7 subscales. Each subscale consists of 5 questions, scored according to the answers to each question: 1. agreement is very much; 2. comparing and agreeing; 3. a little agrees; 4. not to say that; 5. is somewhat different; 6. poor consent; 7. the score is very different, and the score range is 5-35. The score sum of each scale is divided by 35 to give the characteristic index for that scale. Each scale also maps 7 personality respectively (A is adept, B is hesitant to put, C is hesitant, D is regular and swing, E is brief and drifts, F is profound and cautious, G is classic and honorable). And in this embodiment, each personality has a preferred music genre.
Then, step S2 is executed, the data of the vehicle burying points are collected, and the data of the vehicle burying points are randomly divided into a training set and a testing set; the buried point data includes driving information and scene information.
Specifically, referring to fig. 6, in the present embodiment, the driving information includes the driving speed and the driving behavior habit, and the scene information includes the weather condition, the road condition, the holiday information, and the song listening list. The driving information is a continuous variable, and the scene information is a discrete variable.
In this embodiment, the driving information includes driving speed and driving behavior habits, and the scene information includes weather conditions, road conditions, holiday information, and song listening lists. In this embodiment, the driving speed is an average driving speed of the vehicle within 10 minutes from the time of data acquisition. The driving behavior is usually represented by a continuous score S of 0-100, and the value is obtained by counting driving condition records (rapid acceleration and deceleration times, rapid turning times, overspeed times and steering lamp turning times) of hundreds of kilometers before the data acquisition time. For the discrete variable, weather is good and good is represented by 1, and weather is rainy, snowy and foggy is represented by 0; road smoothness is represented by 1 and congestion is represented by 0; holidays are denoted by 1 and weekdays are denoted by 0; and when the song is listened to and the listening time length of each time exceeds more than half of the total time length of the song, expressing the preference degree of the user to the song by using the listening times of the song, and counting to be 0 when the song is not listened to or the listening time length of each time is less than half of the total time length of the song.
More specifically, in the present embodiment, the step S2 of collecting the buried point data of the vehicle includes the steps of:
first, an average traveling speed of the vehicle in a predetermined period of time is obtained, and the average traveling speed is taken as the traveling speed.
Secondly, acquiring the hundred-kilometer running condition of the vehicle before a preset time period, and calculating the driving behavior habit according to the following formula:
wherein S is a driving behavior habit; a. the1The number of rapid acceleration times within hundred kilometers; a. the2The number of rapid deceleration times within hundred kilometers; w is the number of sharp turns in hundred kilometers; l is1Over speed of 10% and less within hundred kilometers; l is2Over 10% and under 20% times within hundred kilometers; l is3Over-speed of 20% and more within hundred kilometers; t is the number of times of turning to the non-turning lamp in hundred kilometers.
It should be noted that, in this embodiment, in order to verify the accuracy of the recommendation algorithm and determine the relevant parameters in the recommendation algorithm, the collected buried point data is calculated according to a ratio of 4: the scale of 1 is randomly divided into a training set and a test set. Specifically, the buried point data algorithm may refer to the prior art, and is not described in detail in this embodiment.
Then, step S3 is executed to pre-classify the driving information of the data of the training set based on the driving information of the buried point data.
In this embodiment, the training data in the training set is pre-classified based on the EM algorithm. Specifically, two feature columns (continuous variables) of the driving speed and the driving behavior habit in the driving scene information are classified, and the driving scene is divided into several categories in advance. In each type of driving scenario, a feature column of one piece of user data includes weather conditions, road conditions, holiday information, and song listening lists. Specifically, the step of performing the EM algorithm may refer to the prior art, which is not described in detail in this embodiment.
Specifically, referring to fig. 2, step S3 includes:
s31: and establishing a Gaussian mixture model according to the driving information of the buried point data.
Specifically, in step S31, when the gaussian mixture model is built according to the driving information, the driving information is used as a feature column, where the feature column of each user to be recommended in each type of driving scene includes scene information.
S32: and determining the maximized model parameters of the Gaussian mixture model by using a maximum expectation algorithm.
Specifically, step S32 includes:
s321: taking the driving speed and the driving behavior habit as sample data of a Gaussian mixture model;
s322: initialization model parameters are set for the gaussian mixture model and S323 and S324 are repeatedly performed.
S323: and calculating the posterior probability of the hidden variables in the Gaussian mixture model according to the initialized model parameters or the last iteration model parameters.
S324: and calculating a likelihood function of the Gaussian mixture model according to the posterior probability, and maximizing the likelihood function to obtain new model parameters.
S325: and when the difference value between the new model parameter and the iterative model parameter is smaller than a preset parameter threshold value, stopping iteration, and taking the new model parameter when iteration is stopped as a maximized model parameter for maximizing the likelihood function.
S33: and pre-classifying the driving information of the data of the training set by utilizing Bayesian information standard.
Specifically, in step S33, the driving scenes of the vehicle are pre-classified according to the following formula:
BIC＝kln(m)-2ln(L)
wherein, BIC is the number of driving scenes of the vehicle; k is the number of the maximized model parameters; m is the number of samples; l is a likelihood function.
It should be noted that, in this embodiment, the category of the driving scene corresponding to the maximum value of the posterior probability of the hidden variable in the gaussian mixture model is the current driving scene of the user to be recommended.
In the present embodiment, the result of the pre-classification is shown in fig. 8 and 9.
Preferably, referring to fig. 3, in the present embodiment, step S3' is further included after step S3.
S3': judging whether the number of songs in a song listening list of a user to be recommended in the current driving scene is smaller than a preset song number threshold value or not in the data of the training set;
if yes, determining a supplementary song with higher similarity to the song in the song listening list from the song library according to the song in the song listening list, and adding the supplementary song to the song listening list;
if not, step S4 is executed.
Specifically, in this embodiment, the threshold of the number of songs is preset to be 10, and of course, those skilled in the art may select other values according to actual needs.
It should be explained that determining the supplementary song with higher similarity to the song in the song listening list from the song library according to the song in the song listening list specifically includes the following steps:
determining a driving scene corresponding to the song listening list, counting the listening situations of songs in the song library of each user in the driving scene, and establishing a total listening list for the song libraries of all the users in the driving scene;
calculating the similarity coefficient of each song in the song listening list of the user to be recommended and other songs which do not exist in the song listening list in the total listening list according to the total listening list;
arranging other songs which do not exist in the song listening list in the total listening list according to the sequence of similarity coefficients from large to small to form a supplementary song list;
and calculating the difference value between the preset song quantity threshold value and the quantity of the songs in the song listening list, and selecting the supplementary songs from the supplementary song list according to the difference value.
Specifically, in an embodiment of the present invention, the method for determining the supplementary song includes: respectively calculating each song listened by the user and the driving range based on the user list of songsJaccard similarity coefficient of other unaccepted songs in song library of scene category, i.e. Jaccard (U)i，Vj) Representing music uiAnd music vjThe correlation of (c). Wherein u isiI ∈ (1, I) represents the ith music the user listens to, uiE.mu, Mu represents a music list listened to by the user in the driving scene, and the music list comprises I songs in total. U shapeiIndicating listening to music u in this driving scenarioiA list of users. v. ofjJ e (1, J) represents the jth music, v, not listened to by the userjE Mv, Mv represents a list of music that the user did not listen to in this driving scenario. VjIndicating listening to music v in this driving scenariojA list of users.
Then, music v is calculated respectivelyjJ e (1, J) and music uiI ∈ (1, I) correlation Jaccard (U)i，Vj) And sum i
To QjAnd J e (1, J) is sorted from large to small, and songs which are not listened to by the user are supplemented into the user listening list according to the sorting order, so that the user listening list at least comprises 10 songs. The number of listening times of the supplemented song is set to 1.
Step S4 is executed next, and the current driving scene of the user to be recommended is determined according to the pre-classification result and the current driving information of the user to be recommended.
Specifically, in this embodiment, an EM algorithm is used to classify two feature columns (continuous variables) of the driving speed and the driving behavior habit in the driving scene information, and based on the model parameter θ of the evaluated gaussian mixture model, a posterior probability p (z | x; θ) of each category to which the recommended user belongs is calculated, where the category z corresponding to the maximum value of the posterior probability is the category to which the recommended user belongs.
Then, step S5 is executed to perform correlation analysis on the scene information corresponding to the current driving scene of the user to be recommended, so as to obtain the music preference information of the user to be recommended.
Specifically, referring to fig. 4, in the present embodiment, step S5 includes:
s51: performing angle cosine correlation analysis on scene information corresponding to the current driving scene of the user to be recommended to acquire music preference information of all users in the current driving scene;
s52: and calculating the music preference similarity of all users in the current driving scene according to the music preference information.
Next, step S6 is executed to determine a plurality of music preference similar users similar to the music preference of the user to be recommended according to the music preference information, and determine a plurality of personality trait similar users similar to the personality trait of the user to be recommended according to the personality trait information.
Specifically, referring to fig. 5, step S6 includes:
s61: determining a plurality of music preference similar users similar to the music preference of the user to be recommended according to the nearest selection strategy; wherein the number of music preference similar users is determined by a cross-validation method.
Specifically, in the embodiment, music preference similar users of the recommended users are found according to the TOP N nearest neighbor selection policy. The parameter N is determined by a 5-fold cross-validation method.
S62: and performing angle cosine correlation analysis on the personality trait information corresponding to the current driving scene of the vehicle to acquire the personality trait information of all users in the current driving scene.
S63: and calculating the personality trait similarity of all users in the current driving scene according to the personality trait information of all users in the current driving scene.
S64: determining a plurality of personality trait similar users similar to the personality traits of the user to be recommended according to the nearest selection strategy; wherein the number of users with similar personality traits is determined by a cross-validation method.
Specifically, in this embodiment, based on the user music personality score (7-dimensional continuous data), the cosine correlation analysis of the included angle is performed, and the similarity of the user music personality is mined. And finding out the music personality similar users of the recommended users according to the TOP N nearest selection strategy. The parameter N is determined by the 5-fold cross-validation method.
Next, step S7 is executed to determine the preference degree of the user to be recommended for each song according to the music preference information of the users with similar music preferences, the personality trait information of the users with similar personality traits, the preference information of the users with similar music preferences for each song, and the preference information of the users with similar personality traits for each song; and constructing a music recommendation list according to the preference of the user to be recommended to each song.
Specifically, in step S7, the preference of the user to be recommended for each song is determined according to the following formula:
wherein, PuiThe preference degree of the user to be recommended to each song is set; sim (u, v) is the music preference similarity of user u and user v; psim (u, t) is the music personality similarity of the user u and the user t; v is a music preference similar user, and t is a personality trait similar user; β is a weight, and the weight β is also determined by the 5-fold cross-validation method.
Then, according to the music recommendation list constructed in step S7, the top n songs of the user are recommended according to the product requirements, sorted by song. Specifically, referring to fig. 10 and fig. 11, in this embodiment, information is acquired according to the listening song list of the user, and the listening song list of the user is supplemented, and finally, a music recommendation list is generated.
It should be further noted that, in this embodiment, the step of performing 5-fold cross validation is as follows:
first, the nearest neighbor N is set to take a value from 5 to 20, and the step size is 1. The weight beta is taken from 0 to 1, and the step length is 0.1. And combining 16 x 11 combinations according to the values of N and beta, and performing 5-fold cross validation on each value combination.
Secondly, the data set is randomly divided into 5 parts, 4 parts of the data set are taken as a training set and 1 part of the data set is taken as a testing set in turn. And recommending according to the values of N and beta, and generating 5 accuracy results through 5 rounds of cross validation.
By adopting the scheme, the driving scene of the vehicle is determined according to the driving information and the scene, and the influence of the driving scene on the mood of the user is fully considered. The method has the advantages that the vehicle driving scenes are pre-classified according to the driving information, when the music recommendation list is constructed subsequently, only the type of the user to be recommended is evaluated according to the pre-classification result, and then the music recommendation list is constructed in the type according to the preference degree and the personality traits, so that the calculation amount is effectively reduced, and the recommendation efficiency is improved.
Further, in the pre-classified driving scenes, scene information such as weather conditions, road conditions, holiday information, and song listening lists, and driving information such as driving speeds and driving behavior habits are considered when calculating the music preference similarity, so that the mood of the user can be sufficiently considered when calculating the music preference.
Furthermore, the music preference similarity, the personality trait similarity and the preferences of the users close to the personality traits of the users to be recommended are combined with corresponding weights, the influence of the mood of the users on the music preference in a real-time driving scene is considered when the music preference similarity is calculated, meanwhile, the personality trait information is considered, and the recommendation accuracy is improved.
While the invention has been shown and described with reference to certain preferred embodiments thereof, it will be understood by those skilled in the art that the foregoing is a more detailed description of the invention, taken in conjunction with the specific embodiments thereof, and that no limitation of the invention is intended thereby. Various changes in form and detail, including simple deductions or substitutions, may be made by those skilled in the art without departing from the spirit and scope of the invention.
Claims (10)
1. A music recommendation method based on a vehicle driving scene is characterized by comprising the following steps:
s1: determining personality trait information of a user to be recommended according to the personality evaluation tool;
s2: collecting data of buried points of a vehicle, and randomly dividing the data of the buried points into a training set and a test set; the buried point data comprises driving information and scene information;
s3: pre-classifying the driving information of the data of the training set according to the driving information of the buried point data;
s4: determining the current driving scene of the user to be recommended according to the pre-classification result and the current driving information of the user to be recommended;
s5: performing relevance analysis on the scene information corresponding to the current driving scene of the user to be recommended to acquire music preference information of the user to be recommended;
s6: determining a plurality of music preference similar users similar to the music preference of the user to be recommended according to the music preference information, and determining a plurality of personality trait similar users similar to the personality traits of the user to be recommended according to the personality trait information;
s7: determining the preference degree of the user to be recommended to each song according to the music preference information of the music preference similar user, the personality trait information of the personality trait similar user, the preference information of the music preference similar user to each song, and the preference information of the personality trait similar user to each song; and constructing a music recommendation list according to the preference of the user to be recommended to each song.
2. The vehicle driving scenario based music recommendation method of claim 1, wherein
In the step S1, the personality assessment tool comprises a sunny music personality assessment scale; and is
In the step S2, the driving information of the buried point data includes a driving speed and a driving behavior habit, and the scene information of the buried point data includes a weather condition, a road condition, holiday information, and a song listening list; wherein
The driving information is a continuous variable, and the scene information is a discrete variable.
3. The vehicle driving scenario-based music recommendation method of claim 2, wherein in step S2, the collecting the buried point data of the vehicle comprises:
obtaining an average running speed of the vehicle in a preset time period, and taking the average running speed as the running speed;
obtaining a hundred kilometer driving condition of the vehicle before the predetermined time period, and calculating the driving behavior habit according to the following formula:
wherein S is a driving behavior habit; a. the1The number of rapid acceleration times within hundred kilometers; a. the2The number of rapid deceleration times within hundred kilometers; w is the number of sharp turns in hundred kilometers; l is1Over speed of 10% and less within hundred kilometers; l is2Over 10% and under 20% times within hundred kilometers; l is3Over-speed of 20% and more within hundred kilometers; t is the number of times of turning to the non-turning lamp in hundred kilometers.
4. The music recommendation method based on vehicle driving scenario of claim 3, wherein step S3 comprises:
s31: establishing a Gaussian mixture model according to the driving information of the buried point data;
s32: determining a maximized model parameter of the Gaussian mixture model by using a maximum expectation algorithm;
s33: pre-classifying the driving information of the data of the training set using Bayesian information criteria.
5. The music recommendation method based on vehicle driving scenes as claimed in claim 4, wherein in step S31, when a gaussian mixture model is built according to the driving information of the buried point data, the driving information is taken as a feature column, wherein the feature column of each user to be recommended in each type of the driving scenes comprises the scene information;
step S32 includes:
s321: taking the driving speed and the driving behavior habit as sample data of the Gaussian mixture model;
s322: setting initialization model parameters for the Gaussian mixture model and repeatedly executing S323 and S324;
s323: calculating the posterior probability of the hidden variable in the Gaussian mixture model according to the initialized model parameter or the last iteration model parameter;
s324: calculating a likelihood function of the Gaussian mixture model according to the posterior probability, and maximizing the likelihood function to obtain new model parameters;
s325: stopping iteration when the difference value between the new model parameter and the iterative model parameter is smaller than a preset parameter threshold value, and taking the new model parameter when iteration is stopped as the maximized model parameter for maximizing the likelihood function;
in step S33, the driving scenes of the vehicle are pre-classified according to the following formula:
BIC＝kln(m)-2ln(L)
wherein, BIC is the number of driving scenes of the vehicle; k is the number of the maximized model parameters; m is the number of samples; l is a likelihood function.
6. The music recommendation method based on vehicle driving scenes according to claim 5, wherein the category of the driving scene corresponding to the maximum value of the posterior probability of the hidden variable in the Gaussian mixture model is the current driving scene of the user to be recommended.
7. The vehicle driving scenario based music recommendation method of claim 6, further comprising, after step S3:
s3': judging whether the number of songs of a song listening list of the user to be recommended in the current driving scene is smaller than a preset song number threshold value or not in the data of the training set;
if yes, determining a supplementary song with higher similarity to the songs in the song listening list from the song library according to the songs in the song listening list, and adding the supplementary song to the song listening list;
if not, step S4 is executed.
8. The music recommendation method based on vehicle driving scenario according to claim 7, wherein in step S3', determining the supplementary songs from the song library with higher similarity to the songs in the song listening list according to the songs in the song listening list comprises:
determining the driving scene corresponding to the song listening list, counting the listening situations of songs in the song library of each user in the driving scene, and establishing a total listening list for the song libraries of all users in the driving scene;
calculating the similarity coefficient of each song in the song listening list of the user to be recommended and other songs which do not exist in the song listening list in the total listening list according to the total listening list;
arranging other songs which do not exist in the song listening list in the total listening list according to the sequence of similarity coefficients from large to small to form a supplementary song list;
and calculating the difference value between the preset song quantity threshold value and the quantity of the songs in the song listening list, and selecting the supplementary songs from the supplementary song list according to the quantity difference value.
9. The vehicle driving scenario-based music recommendation method of claim 8, wherein the step S5 comprises:
s51: performing angle cosine correlation analysis on the scene information corresponding to the current driving scene of the user to be recommended to acquire music preference information of all users in the current driving scene;
s52: calculating music preference similarity of all users in the current driving scene according to the music preference information;
further, the step S6 includes:
s61: determining a plurality of music preference similar users similar to the music preference of the user to be recommended according to the nearest selection strategy; wherein the number of music preference similar users is determined by a cross-validation method;
s62: performing angle cosine correlation analysis on the personality trait information corresponding to the current driving scene of the vehicle to acquire the personality trait information of all users in the current driving scene;
s63: calculating the personality trait similarity of all users in the current driving scene according to the personality trait information of all users in the current driving scene;
s64: determining a plurality of personality trait similar users similar to the personality traits of the user to be recommended according to the nearest selection strategy; wherein the number of users with similar personality traits is determined by a cross-validation method.
10. The music recommendation method based on vehicle driving scenario according to claim 9, wherein in step S7, the preference of the user to be recommended for each song is determined according to the following formula:
wherein, PuiThe preference degree of the user to be recommended to each song is set; sim (u, v) is the music preference similarity of user u and user v; psim (u, t) is the music personality similarity of the user u and the user t; v is a music preference similar user, and t is a personality trait similar user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011298551.2A CN114519122A (en) | 2020-11-19 | 2020-11-19 | Music recommendation method based on vehicle driving scene |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011298551.2A CN114519122A (en) | 2020-11-19 | 2020-11-19 | Music recommendation method based on vehicle driving scene |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114519122A true CN114519122A (en) | 2022-05-20 |
Family
ID=81594341
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011298551.2A Pending CN114519122A (en) | 2020-11-19 | 2020-11-19 | Music recommendation method based on vehicle driving scene |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114519122A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115022363A (en) * | 2022-05-30 | 2022-09-06 | 深圳季连科技有限公司 | Information sharing method considering safety based on Internet of vehicles |
-
2020
- 2020-11-19 CN CN202011298551.2A patent/CN114519122A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115022363A (en) * | 2022-05-30 | 2022-09-06 | 深圳季连科技有限公司 | Information sharing method considering safety based on Internet of vehicles |
| CN115022363B (en) * | 2022-05-30 | 2024-04-16 | 深圳季连科技有限公司 | Information sharing method based on internet of vehicles and considering safety |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110245802B (en) | Cigarette empty-head rate prediction method and system based on improved gradient lifting decision tree | |
| CN107833117B (en) | Bayesian personalized sorting recommendation method considering tag information | |
| US7696427B2 (en) | Method and system for recommending music | |
| CN104731954A (en) | Music recommendation method and system based on group perspective | |
| CN110532379B (en) | Electronic information recommendation method based on LSTM (least Square TM) user comment sentiment analysis | |
| CN111428074B (en) | Audio sample generation method, device, computer equipment and storage medium | |
| KR20080089545A (en) | Information processing device and method, and program | |
| US20090013004A1 (en) | System and Method for the Characterization, Selection and Recommendation of Digital Music and Media Content | |
| CN114117213A (en) | Recommendation model training and recommendation method, device, medium and equipment | |
| EP1530195A2 (en) | Song search system and song search method | |
| Chen | The application of neural network with convolution algorithm in Western music recommendation practice | |
| CN114519122A (en) | Music recommendation method based on vehicle driving scene | |
| CN111177500A (en) | Data object classification method and device, computer equipment and storage medium | |
| Liu et al. | Adaptive music recommendation based on user behavior in time slot | |
| CN116701772B (en) | Data recommendation method and device, computer readable storage medium and electronic equipment | |
| CN108491477B (en) | Neural network recommendation method based on multi-dimensional cloud and user dynamic interest | |
| CN111611432B (en) | Singer classification method based on Labeled LDA model | |
| CN115618024A (en) | Multimedia recommendation method and device and electronic equipment | |
| CN111368131A (en) | User relationship identification method and device, electronic equipment and storage medium | |
| CN114841283A (en) | Method, device, equipment and medium for determining running condition of new energy vehicle | |
| Balkema et al. | Music playlist generation by assimilating GMMs into SOMs | |
| Patwari et al. | Semantically Meaningful Attributes from Co-Listen Embeddings for Playlist Exploration and Expansion. | |
| CN115345656A (en) | Behavior data analysis method for refined marketing | |
| CN115809376A (en) | Intelligent recommendation method based on big teaching data | |
| CN116089713A (en) | Recommendation model training method, recommendation device and computer 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 |