CN116502944A - Live broadcast cargo quality evaluation method based on big data analysis - Google Patents
Live broadcast cargo quality evaluation method based on big data analysis Download PDFInfo
- Publication number
- CN116502944A CN116502944A CN202310376434.0A CN202310376434A CN116502944A CN 116502944 A CN116502944 A CN 116502944A CN 202310376434 A CN202310376434 A CN 202310376434A CN 116502944 A CN116502944 A CN 116502944A
- Authority
- CN
- China
- Prior art keywords
- live broadcast
- live
- feature vector
- acquiring
- commodity
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000013441 quality evaluation Methods 0.000 title claims abstract description 19
- 238000007405 data analysis Methods 0.000 title claims abstract description 18
- 239000013598 vector Substances 0.000 claims abstract description 82
- 238000012706 support-vector machine Methods 0.000 claims abstract description 12
- 230000008451 emotion Effects 0.000 claims description 46
- 230000014509 gene expression Effects 0.000 claims description 35
- 238000012549 training Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 4
- 238000001303 quality assessment method Methods 0.000 claims 10
- 238000010223 real-time analysis Methods 0.000 abstract description 2
- 238000011158 quantitative evaluation Methods 0.000 abstract 2
- 230000006870 function Effects 0.000 description 9
- 230000008909 emotion recognition Effects 0.000 description 7
- 230000002996 emotional effect Effects 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000012512 characterization method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000013145 classification model Methods 0.000 description 2
- 238000013210 evaluation model Methods 0.000 description 2
- 244000144972 livestock Species 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 206010063659 Aversion Diseases 0.000 description 1
- 206010023204 Joint dislocation Diseases 0.000 description 1
- 206010034960 Photophobia Diseases 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 238000011157 data evaluation Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000006397 emotional response Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000008921 facial expression Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000001660 hyperkinetic effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 208000015181 infectious disease Diseases 0.000 description 1
- 206010022000 influenza Diseases 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000036651 mood Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004936 stimulating effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- 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
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Probability & Statistics with Applications (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- General Factory Administration (AREA)
Abstract
The invention discloses a live broadcast and live broadcast quality evaluation method based on big data analysis, which comprises the following steps: acquiring live broadcast cargo carrying data; acquiring a feature vector of a live broadcast attribute based on the live broadcast cargo carrying data; acquiring a preset number of live broadcast on-demand quality classifications based on the live broadcast on-demand data; constructing a live broadcast cargo quality evaluation model based on the feature vector and the live broadcast cargo quality classification; and inputting the feature vector into the live broadcast cargo quality evaluation model to obtain the live broadcast cargo quality category. According to the invention, a support vector machine is adopted to learn big data, a quantitative evaluation model between the live broadcast attribute feature vector and the live broadcast cargo result is established, and the quantitative evaluation model is continuously optimized through real-time analysis and feedback and is used for helping to improve live broadcast cargo quality.
Description
Technical Field
The invention belongs to the field of live broadcast and on-hold goods, and particularly relates to a live broadcast and on-hold goods quality evaluation method based on big data analysis.
Background
With the advent of live delivery, more and more businesses and individuals began selling goods via a live platform. However, different goods require different sales strategies. The live time period, the expression of the host, the sound characteristics and language style of the host can also have important influence on the sales of different types of commodities. However, the current live broadcast hot spot mining method only considers the category and sales data of commodities, and ignores the influence of live broadcast time periods, live broadcast moods of a host broadcast and live broadcast expressions on sales performance. For example, for sports goods, the anchor is recommended to use hyperkinetic, dynamic intonation and rapid speech speed, as well as positive, sun-like emotions; for leisure goods, the anchor is recommended to use a relaxed, mild intonation and slow pace of speech, as well as a relaxed, pleasant emotion. In live delivery, the influence of the expression of the sound and emotion of the host on the live delivery volume is mainly expressed in the following aspects: attracting users' attention, enhancing the characteristics of commodities, and adjusting the live atmosphere.
Emotion transmission refers to expression, infection and sharing of emotion among individuals or groups, and the starting point of transmission is emotion of a sender, so that evaluation and behavioral response of objective things are used as contents, and finally, the emotional response and the transmission behavior of the two parties are caused. In live-broadcast cargo, there are three emotions that are more prominent:
pleasant emotion: the more aggressive the consumer's emotional state, the easier it is to ignore his own needs and create irrational purchase willingness.
Moral emotion: the "moral emotion" is emotion associated with interests or happiness of society or other people, and is emotion generated when an individual obeys or violates moral standards and moral specifications, and mainly comprises aversion, moving emotion, guilt, photophobia, co-emotion, embarrassment, self-luxation and the like.
Emotional performance: the method not only means that the host player purposefully plans and controls the expression of emotion in order to improve the flow rate and commodity sales amount of the live broadcasting room, but also means that the audience performs emotion interaction in order to obtain economic benefits such as prizes, coupons and the like.
The choice of live time period has a tremendous impact on the amount of inventory, you need to know when more people are watching on the spot, how long the host can control the spot in the live room, how long the live will not be tired for the audience, etc. Factors of various aspects are comprehensively considered, and then the live time and duration of each day or week are formulated.
Expression is the most abundant part of human language, and is the reflection of internal emotion. People express the emotion of the mind through expressions such as happiness, anger, grippe, happiness and the like. On the live broadcast and on-goods aspect, the expression plays an important role, and according to related researches, the expression change of the host broadcast has certain relevance to the purchasing desire of consumers during live broadcast.
Disclosure of Invention
In order to solve the technical problems, the invention provides a live broadcast live stock quality evaluation method based on big data analysis, which is used for improving live broadcast live stock quality through real-time analysis and feedback.
In order to achieve the above purpose, the invention provides a live broadcast on-demand quality evaluation method based on big data analysis, which comprises the following steps:
acquiring live broadcast cargo carrying data;
acquiring a feature vector of a live broadcast attribute based on the live broadcast cargo carrying data;
acquiring a preset number of live broadcast on-demand quality classifications based on the live broadcast on-demand data;
constructing a live broadcast cargo quality evaluation model based on the feature vector and the live broadcast cargo quality classification;
and inputting the feature vector into the live broadcast cargo quality evaluation model to obtain the live broadcast cargo quality category.
Optionally, the live tape data includes: the method comprises the steps of sound emotion data of a host, expression data of the host, live time, user comments and commodity sales data.
Optionally, the feature vector includes: sound emotion feature vectors, merchandise feature vectors, live time feature vectors, and expression feature vectors.
Optionally, acquiring the sound emotion feature vector includes:
dividing the live broadcast process into a preset number of time periods, and acquiring sound information and live broadcast carrying capacity of each time period;
carrying out Fourier transform on the sound information and the live broadcast carrying capacity to obtain a sound feature vector;
and acquiring the sound emotion feature vector based on the sound feature vector.
Optionally, obtaining the commodity feature vector includes:
classifying the characteristics of the commodity to obtain the application characteristics and price characteristics of the commodity;
carrying out weight distribution of a preset proportion on the commodities with the application characteristics and the price characteristics;
and carrying out feature coding on the assigned application features and price features to obtain the commodity feature vector.
Optionally, acquiring the live time feature vector includes:
dividing live time into a preset number of time periods;
and acquiring the live time feature vector based on the time period.
Optionally, obtaining the expression feature vector includes:
dividing an image into a preset number of target pixel point areas;
acquiring a central pixel of each target pixel point area;
comparing the central pixel point with the rest pixel points of the target pixel point area to obtain texture information of the central pixel point;
based on the texture information, obtaining expression feature vectors of the target pixel point area,
and acquiring the expression feature vector based on the expression feature vector of the target pixel point area.
Optionally, obtaining the feature vector includes:
and combining and splicing the sound emotion feature vector, the commodity feature vector, the live time feature vector and the expression feature vector to obtain the feature vector.
Optionally, obtaining the preset number of live broadcast on-demand quality classifications includes:
acquiring sales comprehensive data based on the user comments and commodity sales data;
based on the sales comprehensive data, acquiring preset proportion weights of the user comments and commodity sales data;
acquiring the sales comprehensive data sample based on the preset proportion weight;
selecting a preset number of sales comprehensive data samples as a center point, and acquiring the distance between the rest samples and the center point;
and classifying samples based on the distance between the residual samples and the center point, and obtaining the live broadcast belt cargo quality classification.
Optionally, constructing the live-broadcast cargo quality evaluation model includes:
and training a support vector machine by adopting the training set according to the corresponding relation between the live broadcast on-demand quality classification and the feature vector as the training set, and constructing the live broadcast on-demand quality evaluation model.
Compared with the prior art, the invention has the following advantages and technical effects:
according to the invention, the live broadcast with goods effect is improved by establishing a live broadcast with goods evaluation model, the influence of the live broadcast emotion and expression of the host on the live broadcast with goods effect is comprehensively considered, and the accurate live broadcast is realized; according to the invention, an emotion recognition model is adopted to extract the main broadcasting emotion of the live broadcasting audio, then an expression recognition model is used to extract the live broadcasting emotion of the main broadcasting, and the expression characteristics of the live broadcasting emotion are extracted, then the basic attribute and the live broadcasting time period of the commodity are vectorized, a characteristic fusion is carried out and input into a defined classification model, different commodity qualities are distinguished by using an unsupervised clustering K-means algorithm, and the commodity qualities are evaluated by adopting expert scoring, so that a relatively perfect live broadcasting commodity quality evaluation model is finally obtained.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application, illustrate and explain the application and are not to be construed as limiting the application. In the drawings:
fig. 1 is a schematic flow chart of a live broadcast on-demand quality evaluation method based on big data analysis according to an embodiment of the invention.
Detailed Description
It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowcharts, in some cases the steps illustrated or described may be performed in an order other than that illustrated herein.
Examples
The embodiment provides a live broadcast on-demand quality evaluation method based on big data analysis, as shown in fig. 1, which comprises the following steps:
1. acquiring live broadcast cargo carrying data;
collecting live broadcast cargo carrying data: live broadcast and commodity carrying data can be obtained through a live broadcast platform API or other ways, wherein the live broadcast and commodity carrying data comprises sound data of a host broadcast, expression data of the host broadcast, live broadcast time, user comments and commodity sales data.
According to the embodiment, the crawler technology is used for extracting the live broadcast audio with goods of the live broadcast platform, the live broadcast time, user comments, commodity sales data and other related data, and the live broadcast audio on the same day of each anchor is segmented to segment an audio sequence corresponding to each commodity with goods. And obtaining the live expression data of each audio by adopting an expression recognition algorithm for the obtained live audio sequence. For sales data, the amount of goods carried per commodity and user evaluation are of great concern.
2. Acquiring a feature vector of a live broadcast attribute based on live broadcast cargo carrying data;
live attribute feature vector generation: and acquiring live broadcast data in a designated time period, wherein the live broadcast data comprise sound emotion characteristics, commodity characteristics, live broadcast time periods and main broadcasting expressions, and forming feature vectors through normalization. The following methods are adopted for extracting different types of features:
2.1 extraction of emotional characteristics of sound
The emotion feature extraction module is mainly used for extracting emotion features of an input voice signal, and comprises digital preprocessing, endpoint detection and emotion features. The emotion pattern recognition module mainly comprises an emotion recognition model training part and an emotion recognition part, wherein the emotion recognition model training part mainly adopts an SVM method to establish a model for recognizing emotion required, and the emotion recognition part carries out emotion classification on a test voice signal according to the established emotion recognition model so as to realize emotion recognition of a speaker.
Since emotion belongs to the emotion category, the emotion classification can be equal to the voice emotion classification, and the patent mainly focuses on classification of pleasure, moral emotion and emotion performance, and the extracted emotion features are energy and formant features.
Input: live audio sequence
Dividing the live broadcast cargo carrying process into k time periods, and acquiring voice information and live broadcast cargo carrying quantity of each time period.
And carrying out Fourier transform on each sampling sequence, and extracting to obtain a change coefficient to form an audio characteristic. Specifically, for each speech frame, it is windowed and fourier transformed to obtain its power spectrum. The power spectrum is then converted to a Mel frequency spectrum and Discrete Cosine Transformed (DCT) to obtain MFCC coefficients. And connecting all the MFCC coefficients in series to obtain the characteristic vector T of the voice signal.
T＝[t 1 ,t 2 ,t 3 ,t 4 ,t 5 ...t k ]，t k MFCC coefficients expressed as kth frame
The energy characteristics of the voice signal have strong correlation with the expression of emotion. Speech signal energy is typically of both the short-term energy and the short-term average amplitude energy. Because short-time energy is computationally intensive and sensitive to high levels, a short-time average amplitude function is employed here. Assume that the short-time average amplitude function of the nth frame voice signal xn (M) is M n M is then n The estimated expression of (2) is:
n is the total frame number
Calculating a short time average for each frame of each sequenceAverage amplitude energy to obtain short-time average amplitude energy sequence Mm 1 ,m 2 ,m 3 ....m k ]
Coefficient weighting of feature vectors T1T 1 *m 1 ,t 2 *m 2 ,t 3 *m 3 ,...t k *m k ]
And (3) outputting: emotional feature vector T1 of each sampling sequence
2.2 Commodity characterization
The commodity classification is a process of classifying commodities according to a certain classification standard. Generally, the commodity classification may be classified according to different characteristics. The following are features that two consumers prefer to purchase goods:
the application characteristics are as follows: the goods may be classified according to use, such as sports goods, household goods, food, etc.
Price characteristics: goods may be classified by price range, such as high-grade luxury goods, medium-grade goods, low-grade goods, and the like.
When a user purchases a commodity, the user has a certain priority for different commodity characteristics, and can also be said to give different weights to different commodity characteristics when considering that a commodity is purchased. Through big data analysis, the invention adopts weight distribution of 0.6 and 0.4 for two commodity characteristics respectively.
For the purpose feature, one-hot encoding is used for feature encoding. For price characterization, a Min-max normalization is used for characterization normalization. Let minA and maxA be the minimum and maximum value of commodity A price respectively, map an original value x of commodity A into a value x' in interval [0,1] through min-max standardization, its formula is:
x' = (raw data-minimum)/(maximum-minimum)
Splicing the coded characteristic vectors to form a two-dimensional vector T2T 1 ,t 2 ,]Weighting the feature vector by T2T 1 *0.6,t 2 *0.4]Since different features typically have different numerical ranges and scales, the normalization is finally performed once on T2.
2.3 live time period feature
7: 00-10: 00 belongs to the time of bed dependence, the time of the part of people is free, and the income is relatively stable and considerable. At the moment, the number of the anchor persons on the platform is small, the competition is small, and the anchor persons are good opportunities for looping.
12: 00-14: most of the people in the period of 00 noon break are office workers, and the income is stable and considerable. At the moment, the number of the main broadcasters who play on the platform is gradually increased, the competition is gradually increased, the time for maintaining the vermicelli is long, and the office workers are all in noon break, and the people who cannot sleep run the living broadcast room.
19: 00-24: at the time of 00 peak, the platform is in the peak of incoming flow, both the anchor and the audience are in the platform in the period, the local tyrant is more, and the consumption reaches the peak. The start of the war by the anchor robber is the time of stimulating consumption.
We generate a one-dimensional feature vector T3T 1 ]，t 1 E {1,2,3},1, 2,3 correspond to three time periods, respectively.
2.4 expression feature extraction
Because the local binary pattern (Local Binary Pattern, LBP) features cannot be changed obviously due to rotation or illumination change, and the algorithm of the LBP features is easy to calculate, the video facial expression recognition processing can be more effective, and the real-time performance is better. Therefore, the invention adopts LBP operator to extract character expression characteristics. Local refers to the texture of a pixel point on an image, and in many cases refers to the relationship between the point and surrounding pixels; the binary pattern refers to binarization of the center pixel as a threshold.
In the simplest LBP mode, an image is divided into several 3×3 regions in units of pixels. The values of the surrounding 8 pixels are compared with the central pixel, and if not less than the central pixel, the value is set to 1, otherwise 0. Thus, comparing 8 pixel points in the 3×3 range with the central pixel can generate an 8-bit binary number 00010011 (the binary number is converted into a decimal number, namely an LBP code, 256 types are used) to obtain an LBP value of the central pixel point of the window, and the value can reflect the texture information of the region. Obtaining the main broadcasting expression characteristic T4T of each video frame through LBP operator 1 ,t 2 ,t 3 ....t k ]The feature dimension is related to the size of the video frame.
2.5 feature vector stitching
VectorAssembler is a converter that combines a given list of columns into a single column of vectors. In order to train ML models such as logistic regression and decision trees, it is useful to combine the original features and the features generated by the different feature converters into one feature vector. Vector assembler accepts input list types of all numeric types, boolean types, and vector types. In each row, the values of the input columns will be connected into one vector in the order specified.
3. Acquiring a preset number of live broadcast cargo quality classifications based on live broadcast cargo data;
scoring the quality of live broadcast with goods: and combining the commodity carrying capacity of each time period, user comments and an unsupervised clustering model for weight correlation, and classifying the total into 10 types. The professionals were then invited to score these 10 categories, 1 representing poor live tape quality and 10 representing excellent live tape quality.
3.1 data Pre-processing
The current data which we possess comprises commodity carrying capacity and user comments, the user comments need to further mine available information, the obtained user comments are subjected to keyword retrieval, then the obtained keywords are subjected to positive feedback statistics, then the live broadcasting carrying capacity is given with a weight of 0.9, and the obtained positive feedback number is given with a weight of 0.1.
3.2 unsupervised clustering
In this embodiment, the K-Means algorithm is adopted to perform cluster analysis, firstly, we randomly select 10 samples from the sample set as the center points of 10 classes, then look over the distances from the remaining samples to the center points, then divide the distances from other samples to the center points into the 10 classes according to the distance between the other samples and the center points, here we use the euclidean distance as the standard of the distances, calculate the distance from each sample to the 10 center points, then divide the distances according to the distance, for example, the carrying capacity of the commodity a is 300, the positive feedback number obtained from the evaluation is 20, then the sample points obtained after weighting the commodity a are (300×0.9, 20×0.1), and if the center point of the class 1 is the nearest euclidean distance to the center point of the sample, the sample is classified as class 1. Calculating the similarity between two samples is equivalent to calculating the euclidean distance of the two.
3.3 expert scoring
And (5) inviting the live related practitioners to evaluate 10 types of the classified samples, wherein the scores are [1-10 ].
4. Constructing a live broadcast cargo quality evaluation model based on the feature vector and the live broadcast cargo quality classification;
big data evaluation model generation: and according to the corresponding relation between the characteristic vector of each time period and the live broadcast cargo quality of 10 categories, taking the characteristic vector as training data, carrying out model training on big data by adopting a support vector machine, and constructing a live broadcast cargo evaluation model.
4.1 dataset Generation
For 10 classes of live quality, 100 identified samples are given per class as training sets, and 2000 samples (200 each per class) that are not identified are tested.
4.2 Classification model training
The Support Vector Machine (SVM) is a new generation learning algorithm developed on the basis of a statistical learning theory, and the algorithm is well applied to the fields of text classification, handwriting recognition, image classification, bioinformatics and the like. For a classification problem, the SVM builds an optimal hyperplane C0 that distinguishes positive and negative samples in terms of maximum boundaries. In the case of linear inseparable, the SVM uses a kernel function K (P i ,P j ) The feature vector is mapped to a high-dimensional space in which the linear non-separable problem is converted into a linear separable problem.
The basic class-II SVM is expanded, and 10 class-II support vector machines are designed to solve the classification problem of class-10 live broadcast on-demand quality. Wherein K (x) i X) represents an inner product function, and the common inner product functions mainly comprise 3 kinds of polynomial functions, radial basis functions and sigmoid type kernel functions, wherein the radial basis functions are selected in the experimental ring joint.
By establishing a matrix R composed of one-to-one recognition template results, wherein a/b represents live broadcast tape quality category aLive stock with quality b classifier identifies the situation. If there is a feature vector x= (x) 1 ，x 2 ，x 3 ) The 10 output results after passing through the 10 bi-classification SVM classifiers are represented as vectors h (h 1 ,h 2 ,…,h 10 ) And comparing the h vector with each row in the matrix R by adopting the nearest neighbor principle, and obtaining the live broadcast cargo quality category to which the characteristic vector x belongs.
The foregoing is merely a preferred embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions easily conceivable by those skilled in the art within the technical scope of the present application should be covered in the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (10)
1. A method for live broadcast on-demand quality assessment based on big data analysis, comprising:
acquiring live broadcast cargo carrying data;
acquiring a feature vector of a live broadcast attribute based on the live broadcast cargo carrying data;
acquiring a preset number of live broadcast on-demand quality classifications based on the live broadcast on-demand data;
constructing a live broadcast cargo quality evaluation model based on the feature vector and the live broadcast cargo quality classification;
and inputting the feature vector into the live broadcast cargo quality evaluation model to obtain the live broadcast cargo quality category.
2. A method of live tape quality assessment based on big data analysis as claimed in claim 1, wherein the live tape data comprises: the method comprises the steps of sound emotion data of a host, expression data of the host, live time, user comments and commodity sales data.
3. A method of live tape quality assessment based on big data analysis according to claim 1, wherein the feature vector comprises: sound emotion feature vectors, merchandise feature vectors, live time feature vectors, and expression feature vectors.
4. A method of live tape quality assessment based on big data analysis according to claim 3, wherein obtaining the sound emotion feature vector comprises:
dividing the live broadcast process into a preset number of time periods, and acquiring sound information and live broadcast carrying capacity of each time period;
carrying out Fourier transform on the sound information and the live broadcast carrying capacity to obtain a sound feature vector;
and acquiring the sound emotion feature vector based on the sound feature vector.
5. A method of live tape quality assessment based on big data analysis according to claim 3, wherein obtaining the commodity feature vector comprises:
classifying the characteristics of the commodity to obtain the application characteristics and the price characteristics of the commodity;
carrying out weight distribution of a preset proportion on the application characteristics and the price characteristics of the commodity;
and carrying out feature coding on the assigned application features and price features to obtain the commodity feature vector.
6. A method of live tape quality assessment based on big data analysis according to claim 3, wherein obtaining the live time feature vector comprises:
dividing live time into a preset number of time periods;
and acquiring the live time feature vector based on the time period.
7. The method of claim 3, wherein obtaining the expression feature vector comprises:
dividing an image into a preset number of target pixel point areas;
acquiring a central pixel of each target pixel point area;
comparing the central pixel point with the rest pixel points of the target pixel point area to obtain texture information of the central pixel point;
based on the texture information, obtaining expression feature vectors of the target pixel point area,
and acquiring the expression feature vector based on the expression feature vector of the target pixel point area.
8. A method of live tape quality assessment based on big data analysis according to claim 3, wherein obtaining the feature vector comprises:
and combining and splicing the sound emotion feature vector, the commodity feature vector, the live time feature vector and the expression feature vector to obtain the feature vector.
9. The method for live tape quality assessment based on big data analysis of claim 2, wherein obtaining a preset number of live tape quality classifications comprises:
acquiring sales comprehensive data based on the user comments and the commodity sales data;
based on the sales comprehensive data, acquiring preset proportion weights of the user comments and the commodity sales data;
acquiring the sales comprehensive data sample based on the preset proportion weight;
selecting a preset number of sales comprehensive data samples as a center point, and acquiring the distance between the rest samples and the center point;
and classifying samples based on the distance between the residual samples and the center point, and obtaining the live broadcast belt cargo quality classification.
10. The method for live tape quality assessment based on big data analysis of claim 9, wherein constructing the live tape quality assessment model comprises:
and training a support vector machine by adopting a training set according to the corresponding relation between the live broadcast on-demand quality classification and the feature vector as the training set, and constructing the live broadcast on-demand quality evaluation model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310376434.0A CN116502944A (en) | 2023-04-10 | 2023-04-10 | Live broadcast cargo quality evaluation method based on big data analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310376434.0A CN116502944A (en) | 2023-04-10 | 2023-04-10 | Live broadcast cargo quality evaluation method based on big data analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116502944A true CN116502944A (en) | 2023-07-28 |
Family
ID=87317551
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310376434.0A Pending CN116502944A (en) | 2023-04-10 | 2023-04-10 | Live broadcast cargo quality evaluation method based on big data analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116502944A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111754107A (en) * | 2020-06-22 | 2020-10-09 | 广州虎牙科技有限公司 | Anchor value evaluation method and device, electronic equipment and readable storage medium |
| CN113780217A (en) * | 2021-09-16 | 2021-12-10 | 中国平安人寿保险股份有限公司 | Live broadcast auxiliary prompting method and device, computer equipment and storage medium |
| WO2022001349A1 (en) * | 2020-07-01 | 2022-01-06 | 北京京东振世信息技术有限公司 | Method and device for information analysis |
| CN115271335A (en) * | 2022-06-15 | 2022-11-01 | 同芙集团(中国)股份有限公司 | Live broadcast tape goods owner broadcast evaluation method, device, equipment and storage medium |
| CN115630869A (en) * | 2022-10-31 | 2023-01-20 | 合肥工业大学 | Comprehensive evaluation method for E-commerce anchor capability based on entropy weight method |
| CN115713257A (en) * | 2022-11-07 | 2023-02-24 | 浙江工商大学 | Anchor expressive force evaluation method and device based on multi-mode fusion and computing equipment |
| CN115879972A (en) * | 2022-03-14 | 2023-03-31 | 杭州鸭梨互动网络科技有限公司 | Data processing method for oriented analysis and prediction of e-commerce platform |
-
2023
- 2023-04-10 CN CN202310376434.0A patent/CN116502944A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111754107A (en) * | 2020-06-22 | 2020-10-09 | 广州虎牙科技有限公司 | Anchor value evaluation method and device, electronic equipment and readable storage medium |
| WO2022001349A1 (en) * | 2020-07-01 | 2022-01-06 | 北京京东振世信息技术有限公司 | Method and device for information analysis |
| CN113780217A (en) * | 2021-09-16 | 2021-12-10 | 中国平安人寿保险股份有限公司 | Live broadcast auxiliary prompting method and device, computer equipment and storage medium |
| CN115879972A (en) * | 2022-03-14 | 2023-03-31 | 杭州鸭梨互动网络科技有限公司 | Data processing method for oriented analysis and prediction of e-commerce platform |
| CN115271335A (en) * | 2022-06-15 | 2022-11-01 | 同芙集团(中国)股份有限公司 | Live broadcast tape goods owner broadcast evaluation method, device, equipment and storage medium |
| CN115630869A (en) * | 2022-10-31 | 2023-01-20 | 合肥工业大学 | Comprehensive evaluation method for E-commerce anchor capability based on entropy weight method |
| CN115713257A (en) * | 2022-11-07 | 2023-02-24 | 浙江工商大学 | Anchor expressive force evaluation method and device based on multi-mode fusion and computing equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kächele et al. | Fusion of audio-visual features using hierarchical classifier systems for the recognition of affective states and the state of depression | |
| US8510252B1 (en) | Classification of inappropriate video content using multi-scale features | |
| CN109614895A (en) | A method of the multi-modal emotion recognition based on attention Fusion Features | |
| WO2020206066A1 (en) | Method for optimizing media and marketing content using cross-platform video intelligence | |
| CN105047194B (en) | A kind of self study sound spectrograph feature extracting method for speech emotion recognition | |
| Senthilkumar et al. | Speech emotion recognition based on Bi-directional LSTM architecture and deep belief networks | |
| CN113158023A (en) | Public digital life accurate classification service method based on mixed recommendation algorithm | |
| Heng et al. | Hybrid supervised deep learning for ethnicity classification using face images | |
| CN110992988A (en) | Speech emotion recognition method and device based on domain confrontation | |
| CN115393933A (en) | Video face emotion recognition method based on frame attention mechanism | |
| Boishakhi et al. | Multi-modal hate speech detection using machine learning | |
| Lin et al. | Meta-classification of multimedia classifiers | |
| CN115101077A (en) | Voiceprint detection model training method and voiceprint recognition method | |
| Mujaddidurrahman et al. | Speech Emotion Recognition Using 2D-CNN with Data Augmentation | |
| Wu et al. | Ordinal Triplet Loss: Investigating Sleepiness Detection from Speech. | |
| Boychuk et al. | An exploratory sentiment and facial expressions analysis of data from photo-sharing on social media: The case of football violence | |
| Hadid | Analyzing facial behavioral features from videos | |
| CN113139525A (en) | Multi-source information fusion-based emotion recognition method and man-machine interaction system | |
| CN116502944A (en) | Live broadcast cargo quality evaluation method based on big data analysis | |
| Ramashini et al. | A Novel Approach of Audio Based Feature Optimisation for Bird Classification. | |
| Doshi et al. | Acoustic Scene Analysis and Classification Using Densenet Convolutional Neural Network | |
| CN112069897A (en) | Knowledge graph-based voice and micro-expression recognition suicide emotion sensing method | |
| Primorac et al. | Audio-visual biometric recognition via joint sparse representations | |
| Meghjani et al. | Bimodal information analysis for emotion recognition | |
| WO2016063403A1 (en) | Data analysis system, data analysis method, and data analysis program |
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 |