CN115842812B - User perception evaluation method and system based on PCA and integrated self-encoder - Google Patents
User perception evaluation method and system based on PCA and integrated self-encoder Download PDFInfo
- Publication number
- CN115842812B CN115842812B CN202211452533.4A CN202211452533A CN115842812B CN 115842812 B CN115842812 B CN 115842812B CN 202211452533 A CN202211452533 A CN 202211452533A CN 115842812 B CN115842812 B CN 115842812B
- Authority
- CN
- China
- Prior art keywords
- data
- index
- user
- value
- encoder
- 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.)
- Active
Links
- 230000008447 perception Effects 0.000 title claims abstract description 72
- 238000011156 evaluation Methods 0.000 title claims abstract description 23
- 238000000034 method Methods 0.000 claims abstract description 36
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000012545 processing Methods 0.000 claims abstract description 18
- 238000010606 normalization Methods 0.000 claims abstract description 13
- 230000004044 response Effects 0.000 claims abstract description 11
- 238000004140 cleaning Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 26
- 238000004364 calculation method Methods 0.000 claims description 16
- 208000025779 tropical pancreatitis Diseases 0.000 claims description 16
- 238000005070 sampling Methods 0.000 claims description 15
- 238000003860 storage Methods 0.000 claims description 12
- 238000009826 distribution Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 8
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 8
- 230000009747 swallowing Effects 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 5
- 238000012935 Averaging Methods 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 6
- 238000011161 development Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- 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
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Telephonic Communication Services (AREA)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
Abstract
The invention discloses a user perception evaluation method and a system based on PCA and an integrated self-encoder, which belong to the technical field of artificial intelligence and user perception evaluation, and the technical problem to be solved by the invention is that the robustness is insufficient, the user perception evaluation task under different environments and different service types is difficult to adapt, and the adopted technical scheme is as follows: the method comprises the following steps: acquiring a voice-related KQI index and a video-related KQI index of a high-speed rail user through a sensor; data cleaning: when a null value exists under the VOLTE response call drop rate index of one piece of data, setting the null value to zero; when the other arbitrary indexes in one piece of data have blank values, deleting the piece of user data and not participating in modeling; preprocessing the cleaned data; acquiring initial perception by using PCA, and acquiring component weights by normalization processing; training an integrated self-encoder; obtaining a reconstruction weight through a reconstruction error of the integrated self-encoder; a user perception score is obtained.
Description
Technical Field
The invention relates to the technical field of artificial intelligence and user perception evaluation, in particular to a user perception evaluation method and system based on PCA and an integrated self-encoder.
Background
With rapid development of wireless communication technology, cost-effective intelligent terminals are becoming popular. The behaviors of the end users show a development trend of diversification and complexity, and the service types are also more and more abundant. By evaluating the perception of the user, the perception situation of the user can be known, and the network link planning, the network link optimization, the market development and the like can be accurately carried out to provide more accurate references, so that an operator can schedule resources more efficiently to provide better quality service for the user, and the terminal user has better user perception.
Currently, the score evaluation method for user perception usually enables an expert to formulate a unified evaluation standard to evaluate the user perception of all service types based on own experience. This evaluation criterion is fixed unless the expert reformulates a new evaluation criterion. Since the distribution of data of different service types is different, even the distribution of data of the same service type in different times is also different to some extent. Meanwhile, the terminal user has different perception standards for different service types, so that the current evaluation method cannot adaptively obtain accurate user perception evaluation results aiming at different service types.
Therefore, the existing score evaluation method for user perception is insufficient in robustness, and is difficult to adapt to user perception evaluation tasks under different environments and different service types.
Disclosure of Invention
The technical task of the invention is to provide a user perception evaluation method and a system based on PCA and an integrated self-encoder, which solve the problem that the robustness is insufficient and the user perception evaluation task under different environments and different service types is difficult to adapt.
The technical task of the invention is realized in the following way, namely a user perception evaluation method based on PCA and an integrated self-encoder, which comprises the following steps:
acquiring a voice-related KQI index and a video-related KQI index of a high-speed rail user through a sensor;
data cleaning: when a null value exists under the VOLTE response call drop rate index of one piece of data, setting the null value to zero; when the other arbitrary indexes in one piece of data have blank values, deleting the piece of user data and not participating in modeling;
preprocessing the cleaned data: the z-score method is used for carrying out standardization processing, standardized data are used for weight calculation, and influence of the dimension of the index on weight calculation is reduced;
acquiring initial perception by using PCA, and acquiring component weights by normalization processing;
training an integrated self-encoder;
obtaining a reconstruction weight through a reconstruction error of the integrated self-encoder;
obtaining a user perception score: and obtaining an average value of the component weights and the reconstruction weights as index weights, obtaining user scores according to the data of each user and the data value of each index, and multiplying the user scores of each index with the corresponding index weights to obtain user perception scores.
Preferably, the KQI indexes related to the video comprise uplink average RTTms, downlink average RTTms, stream media XKB starting time delay, stream media pause ratio, stream media pause frequency TIMESMIN, uplink TCP disorder rate, downlink TCP disorder rate, uplink TCP retransmission rate, downlink TCP retransmission rate and stream media downlink rate kbps;
the voice related KQI index VOLTE initial call network call completing rate, VOLTE response call dropping rate, uplink average MOS, downlink average MOS, uplink RTP packet dropping rate, downlink RTP packet dropping rate, uplink single pass proportion, downlink single pass proportion, uplink word swallowing proportion, downlink word swallowing proportion, uplink intermittent proportion and downlink intermittent proportion.
Preferably, the normalization treatment using the z-score method is specifically as follows:
obtaining a mean value and a standard value;
the data for each index is subtracted from the corresponding mean value and divided by the standard value so that the data meets the standard normal distribution.
Preferably, the initial perception is obtained using PCA, and the component weights are obtained by normalization processing as follows:
the voice component weight and the video component weight of the user are independent, and the voice component weight and the video component weight are obtained separately;
sequentially calculating component weights for video data and voice data: respectively calculating covariance matrixes of video data and voice data of a user, and carrying out eigenvalue decomposition on the covariance matrixes to obtain n eigenvalues mu= { mu 1 ,μ 2 ,...,μ n Sum n eigenvectors v= { v 1 ,v 2 ,...,v n }；
Dividing each eigenvalue by the sum of all eigenvalues to obtain a corresponding variance contribution ratio sigma i The formula is as follows:
multiplying each variance contribution rate by all components of the corresponding feature vector yields a weighted feature vector v' as follows:
v′＝{v 1 ×σ 1 ,...,v n ×σ n }；
summing each row of the feature vector v' results in a set of weighted values w, the formula being as follows:
w＝{w 1 ,w 2 ,...,w n }；
and normalizing the weighted values to obtain component weight, wherein the formula is as follows:
weight＝{w 1 ′,w′ 2 ,...,w′ n }。
preferably, the training integrated self-encoder is specifically as follows:
normal value selection: the self-encoder is used for detecting abnormal values, and training data which are all normal values are needed;
because the data approximately accords with normal distribution, a 3 sigma method is adopted for selecting normal values;
calculating the mean value theta of the data of the ith index i And standard deviation sigma i Will have a value of θ i ±3×σ i Data within (mean plus three standard deviations) are identified as normal values;
randomly sampling all normal values with replacement, and obtaining n normal values each time to form a sample until a moderate-scale data set is obtained;
training a self-encoder by using the constructed normal value to reconstruct samples, and establishing a mapping between original samples and reconstructed samples of the normal value;
a self-encoder trained with normal values will have difficulty reconstructing the outlier samples, one for each index.
Preferably, the reconstruction weights are obtained by integrating the reconstruction errors of the self-encoder as follows:
reservoir sampling is carried out on the user data, n pieces of user data are obtained through each sampling, and an n multiplied by f user data matrix A is obtained; wherein f represents the number of features;
transpose the user data matrix A to obtain an f×n matrix A T ；
The total is sampled m times to obtain a f multiplied by m multiplied by n user matrix B;
the method comprises the steps of importing a self-encoder in pre-training, calculating reconstruction errors of m samples for each index, and solving an average value as a single sampling reconstruction error of the index;
and repeating the reconstruction error operation for h times by adopting the bagging concept of the integrated algorithm to obtain a matrix C of h multiplied by f, and averaging the columns of the matrix C to obtain the reconstruction weight.
More preferably, the user perception score is obtained as follows:
for a forward index, i.e., an index whose score should be higher the larger the value, the formula is as follows:
wherein x is the data value of each index; a and B are the lowest threshold and the highest threshold of the score respectively; if the data value is lower than A, the index score is 0; if the data value is higher than B, the index score is 100; the data value is between A and B, thenScoring is performedCalculating;
for negative indicators, i.e. indicators where the score should be higher the smaller the value, the formula is as follows:
wherein x is the data value of each index; a and B are the lowest threshold and the highest threshold of the score respectively; if the data value is higher than A, the index score is 0; if the data value is lower than B, the index score is 100; the data value is between A and B, thenPerforming score calculation;
the user score of each index is multiplied by the corresponding index weight to obtain a user perception score.
A PCA-based and integrated self-encoder based user perception assessment system, the system comprising,
the index acquisition module is used for acquiring a KQI index related to the voice of the high-speed rail user and a KQI index related to the video through the sensor;
the data cleaning module is used for cleaning data and specifically comprises the following steps: when a null value exists under the VOLTE response call drop rate index of one piece of data, setting the null value to zero; when the other arbitrary indexes in one piece of data have blank values, deleting the piece of user data and not participating in modeling;
the pretreatment module is used for carrying out pretreatment on the cleaned data, and specifically comprises the following steps: the z-score method is used for carrying out standardization processing, standardized data are used for weight calculation, and influence of the dimension of the index on weight calculation is reduced;
the component weight acquisition module is used for acquiring initial perception by using PCA and acquiring component weights through normalization processing;
the training module is used for training the integrated self-encoder;
the reconstruction weight acquisition module is used for acquiring reconstruction weights through reconstruction errors of the integrated self-encoder;
the user perception score acquisition module is used for solving the mean value of the component weights and the reconstruction weights to serve as index weights, acquiring user scores according to the data of each user and the data value of each index, and multiplying the user scores of each index with the corresponding index weights to obtain the user perception scores.
An electronic device, comprising: a memory and at least one processor;
wherein the memory has a computer program stored thereon;
the at least one processor executes the computer program stored by the memory, causing the at least one processor to perform the PCA and integrated self-encoder based user perception assessment method as described above.
A computer readable storage medium having stored therein a computer program executable by a processor to implement a PCA and integrated self-encoder based user perception assessment method as described above.
The user perception evaluation method and system based on PCA and integrated self-encoder of the invention has the following advantages:
firstly, the PCA and the integrated self-encoder are used for calculating the user perception weight, and the relevance between indexes and the stability of data in the indexes can be considered simultaneously in the process of modeling the weight, so that the user perception can be evaluated more accurately;
according to the invention, the user perception data is obtained through the sensor, the data is cleaned, the data is standardized, the PCA is used for initial perception calculation, the integrated self-encoder is used for training, the integrated self-encoder is used for calculating the reconstruction error and calculating the user perception score, the robustness of the user perception score evaluation is improved, and the user perception evaluation task under different environments and different service types can be adapted.
Drawings
The invention is further described below with reference to the accompanying drawings.
FIG. 1 is a flow chart diagram of a user perception assessment method based on PCA and integrated self-encoders;
fig. 2 is a schematic diagram of the KQI index of the high-speed rail user.
Detailed Description
The user perception assessment method and system based on PCA and integrated self-encoder of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
Example 1:
as shown in fig. 1, the present embodiment provides a user perception evaluation method based on PCA and integrated self-encoder, which specifically includes the following steps:
s1, acquiring a KQI index related to voice and a KQI index related to video of a high-speed rail user through a sensor;
s2, data cleaning: because some high-speed rail users may only watch video or only make voice calls when riding high-speed rail users, users with different video and voice data need to be screened out; when a null value exists under the VOLTE response call drop rate index of one piece of data, setting the null value to zero; when the other arbitrary indexes in one piece of data have blank values, deleting the piece of user data and not participating in modeling;
s3, preprocessing the cleaned data: the z-score method is used for carrying out standardization processing, standardized data are used for weight calculation, and influence of the dimension of the index on weight calculation is reduced;
s4, acquiring initial perception by using PCA, and acquiring component weights through normalization processing;
s5, training an integrated self-encoder;
s6, acquiring a reconstruction weight through a reconstruction error of the integrated self-encoder;
s7, obtaining a user perception score: and obtaining an average value of the component weights and the reconstruction weights as index weights, obtaining user scores according to the data of each user and the data value of each index, and multiplying the user scores of each index with the corresponding index weights to obtain user perception scores.
As shown in fig. 1, the KQI indexes related to the video in step S1 of the present embodiment include an uplink average RTTms, a downlink average RTTms, a stream XKB start delay, a stream pause duty ratio, a stream pause frequency tinsmin, an uplink TCP disorder rate, a downlink TCP disorder rate, an uplink TCP retransmission rate, a downlink TCP retransmission rate, and a stream downlink rate kbps;
the voice related KQI index VOLTE initial call network call completing rate, VOLTE response call dropping rate, uplink average MOS, downlink average MOS, uplink RTP packet dropping rate, downlink RTP packet dropping rate, uplink single pass proportion, downlink single pass proportion, uplink word swallowing proportion, downlink word swallowing proportion, uplink intermittent proportion and downlink intermittent proportion.
The normalization processing using the z-score method in step S3 of this embodiment is specifically as follows:
s301, acquiring a mean value and a standard value;
s302, subtracting the corresponding mean value from the data of each index and dividing the data by a standard value to enable the data to accord with standard normal distribution.
The initial perception is obtained by using PCA in step S4 of this embodiment, and the component weights are obtained by normalization processing as follows:
s401, the voice component weight and the video component weight of the user are independent, and the voice component weight and the video component weight are obtained separately;
s402, sequentially calculating component weights for video data and voice data: respectively calculating covariance matrixes of video data and voice data of a user, and carrying out eigenvalue decomposition on the covariance matrixes to obtain n eigenvalues mu= { mu 1 ,μ 2 ,...,μ n Sum n eigenvectors v= { v 1 ,v 2 ,...,v n }；
S403, dividing each characteristic value by the sum of all characteristic values to obtain a corresponding variance contribution rate sigma i The formula is as follows:
s404, multiplying each variance contribution rate by all components of the corresponding eigenvector to obtain a weighted eigenvector v', wherein the formula is as follows:
v′＝{v 1 ×σ 1 ,...,v n ×σ n }；
s405, summing each row of the feature vector v' to obtain a set of weighted values w, wherein the formula is as follows:
w＝{w 1 ,w 2 ,...,w n }；
s406, normalizing the weighted value to obtain a component weight, wherein the formula is as follows:
weight＝{w 1 ′,w′ 2 ,...,w′ n }。
the training integrated self-encoder in step S5 of the present embodiment is specifically as follows:
s501, normal value selection: the self-encoder is used for detecting abnormal values, and training data which are all normal values are needed;
s502, selecting a normal value by adopting a 3 sigma method because the data approximately accords with normal distribution;
s503, calculating the average value theta of the data of the ith index i And standard deviation sigma i Will have a value of θ i ±3×σ i Data within (mean plus three standard deviations) are identified as normal values;
s504, randomly sampling all normal values with replacement, and obtaining n normal values each time to form a sample until a data set with moderate scale is obtained;
s505, training a self-encoder by using the constructed normal value to reconstruct samples, and establishing the mapping between the original samples and reconstructed samples of the normal value;
s506, the self-encoder trained by the normal value is difficult to reconstruct abnormal samples, and each index is stored in a self-encoder after training.
The reconstruction weights obtained by integrating the reconstruction errors of the self-encoder in step S6 of the present embodiment are specifically as follows:
s601, carrying out reservoir sampling on user data, and obtaining n pieces of user data in each sampling to obtain an n multiplied by f user data matrix A; wherein f represents the number of features;
s602, transpose the user data matrix A to obtain an f×n matrix A T ；
S603, performing m times of sampling operation to obtain an f multiplied by m multiplied by n user matrix B;
s604, importing the error into a pre-trained self-encoder, calculating the reconstruction errors of m samples for each index, and solving the average value as a single sampling reconstruction error of the index;
s605, adopting a bagging concept of an integration algorithm, repeating the reconstruction error operation for h times to obtain a matrix C of h multiplied by f, and averaging the columns of the matrix C to obtain the reconstruction weight.
The obtaining of the user perception score in step S7 of this embodiment is specifically as follows:
s701, for the forward index, i.e. the index whose score should be higher as the numerical value is larger, the formula is as follows:
wherein x is the data value of each index; a and B are the lowest threshold and the highest threshold of the score respectively; if the data value is lower than A, the index score is 0; if the data value is higher than B, the index score is 100; the data value is between A and B, thenPerforming score calculation;
s702, for negative indicators, i.e. indicators with smaller values and higher scores, the formula is as follows:
wherein x is the data value of each index; a and B are the lowest threshold and the highest threshold of the score respectively; if the data value is higher than A, the index score is 0; if the data value is lower than B, the index score is 100; the data value is between A and B, thenPerforming score calculation;
s703, multiplying the user score of each index by the corresponding index weight to obtain a user perception score.
Example 2:
the present embodiment provides a PCA and integrated self-encoder based user perception assessment system including,
the index acquisition module is used for acquiring a KQI index related to the voice of the high-speed rail user and a KQI index related to the video through the sensor; wherein, the KQI index related to the voice and the KQI index related to the video of the high-speed rail user are obtained through the sensor. For video data, data with 10 indexes, namely, uplink average RTTms, downlink average RTTms, streaming media XKB starting time delay, streaming media pause duty ratio, streaming media pause frequency TIMEMIN, uplink TCP disorder rate, downlink TCP disorder rate, uplink TCP retransmission rate, downlink TCP retransmission rate and streaming media downlink rate kbps, are required to be collected. For voice data, the data of 12 indexes of VOLTE initial call network call completing rate, VOLTE response dropped call rate, uplink average MOS, downlink average MOS, uplink RTP packet loss rate, downlink RTP packet loss rate, uplink single-pass proportion, downlink single-pass proportion, uplink word-swallowing proportion, downlink word-swallowing proportion, uplink intermittent proportion and downlink intermittent proportion are required to be collected.
The data cleaning module is used for cleaning data and specifically comprises the following steps: because some high-speed rail users may only watch video or only make voice calls when riding high-speed rail users, users with different video and voice data need to be screened out; when a null value exists under the VOLTE response call drop rate index of one piece of data, setting the null value to zero; when the other arbitrary indexes in one piece of data have blank values, deleting the piece of user data and not participating in modeling;
the pretreatment module is used for carrying out pretreatment on the cleaned data, and specifically comprises the following steps: and (3) carrying out standardization processing on the cleaned data, and carrying out standardization on the data of each index by using a z-score method. The mean and standard values are calculated first. The data for each index is subtracted from the corresponding mean and divided by the standard deviation such that the data meets the standard normal distribution. Reducing the influence of the dimension of the index on the weight calculation, wherein the standardized data is used for the weight calculation;
the component weight acquisition module is used for acquiring initial perception by using PCA and acquiring component weights through normalization processing; the method comprises the following steps: the user's speech component weights and video component weights are independent of each other and are calculated separately. Sequentially calculating component weights for video data and voice data: and calculating a covariance matrix of the data, and carrying out eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors. And calculating a variance contribution value of each characteristic value, and carrying out normalization processing on the variance contribution value to obtain a variance contribution rate. Multiplying each variance contribution by all components of the corresponding feature vector yields a weighted feature vector. All components of each weighted feature vector are summed to obtain a set of weighted values. And normalizing the weighted value to obtain the component weight.
The training module is used for training the integrated self-encoder; the method comprises the following steps: firstly, normal value selection is carried out, and training data which are all normal values are needed for abnormal value detection by a self-encoder. Because the data approximately accords with normal distribution, the selection of normal values is carried out by adopting a 3-method. The mean and standard deviation of the data of the first index are calculated, and the data whose value is within (the mean plus three times the standard deviation) are identified as normal values. All normal values were sampled randomly with a return, and each normal value was taken to form a sample until a moderate-scale dataset was obtained. The self-encoder is trained with the constructed dataset for sample reconstruction, creating a mapping of the original samples and reconstructed samples of normal values. A self-encoder trained with normal values will have difficulty reconstructing the outlier samples, one for each index.
The reconstruction weight acquisition module is used for acquiring reconstruction weights through reconstruction errors of the integrated self-encoder; the method comprises the following steps: and carrying out reservoir sampling on the user data, and obtaining n pieces of user data for each sampling to obtain an n multiplied by f user data matrix A, wherein f represents the number of the features. Transpose matrix A to obtain an f×n matrix A T . The total is sampled m times to obtain a f x m x n user matrix B. A pre-trained self-encoder is introduced in step S5, for each index, the reconstruction error of the m samples is calculated and the mean is taken as the single sample reconstruction error for that index. Adopting the concept of the integrated algorithm to perform the above operationsRepeating for h times to obtain a matrix C of h multiplied by f, and obtaining a reconstruction weight by averaging the columns of the matrix C.
The user perception score acquisition module is used for solving the mean value of the component weights and the reconstruction weights to serve as index weights, acquiring user scores according to the data of each user and the data value of each index, and multiplying the user scores of each index with the corresponding index weights to obtain the user perception scores. For data values below or above the preset index range, a score of 0 or 100 is directly assigned. The user score of each index is multiplied by the corresponding index weight and summed to obtain the user perception score.
Example 3:
the embodiment of the invention also provides electronic equipment, which comprises: a memory and a processor;
wherein the memory stores computer-executable instructions;
the processor executes the computer-executable instructions stored by the memory, causing the processor to perform the PCA and integrated self-encoder based user perception assessment method in any of the embodiments of the present invention.
The processor may be a Central Processing Unit (CPU), but may also be other general purpose processors, digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory may be used to store computer programs and/or modules, and the processor implements various functions of the electronic device by running or executing the computer programs and/or modules stored in the memory, and invoking data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function, and the like; the storage data area may store data created according to the use of the terminal, etc. The memory may also include high-speed random access memory, but may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, memory card only (SMC), secure Digital (SD) card, flash memory card, at least one disk storage period, flash memory device, or other volatile solid state memory device.
Example 4:
embodiments of the present invention also provide a computer readable storage medium having stored therein a plurality of instructions that are loaded by a processor to cause the processor to perform the PCA and integrated self-encoder based user perception assessment method of any of the embodiments of the present invention. Specifically, a system or apparatus provided with a storage medium on which a software program code realizing the functions of any of the above embodiments is stored, and a computer (or CPU or MPU) of the system or apparatus may be caused to read out and execute the program code stored in the storage medium.
In this case, the program code itself read from the storage medium may realize the functions of any of the above-described embodiments, and thus the program code and the storage medium storing the program code form part of the present invention.
Examples of storage media for providing program code include floppy disks, hard disks, magneto-optical disks, optical disks (e.g., CD-ROMs, CD-R, CD-RWs, DVD-ROMs, DVD-RYM, DVD-RWs, DVD+RWs), magnetic tapes, nonvolatile memory cards, and ROMs. Alternatively, the program code may be downloaded from a server computer by a communication network.
Further, it should be apparent that the functions of any of the above-described embodiments may be implemented not only by executing the program code read out by the computer, but also by causing an operating system or the like operating on the computer to perform part or all of the actual operations based on the instructions of the program code.
Further, it is understood that the program code read out by the storage medium is written into a memory provided in an expansion board inserted into a computer or into a memory provided in an expansion unit connected to the computer, and then a CPU or the like mounted on the expansion board or the expansion unit is caused to perform part and all of actual operations based on instructions of the program code, thereby realizing the functions of any of the above embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (7)
1. The user perception evaluation method based on PCA and integrated self-encoder is characterized by comprising the following specific steps:
acquiring a voice-related KQI index and a video-related KQI index of a high-speed rail user through a sensor;
data cleaning: when a null value exists under the VOLTE response call drop rate index of one piece of data, setting the null value to zero; when the other arbitrary indexes in one piece of data have blank values, deleting the piece of user data and not participating in modeling;
preprocessing the cleaned data: performing standardization processing by using a z-score method, wherein standardized data are used for weight calculation;
acquiring initial perception by using PCA, and acquiring component weights by normalization processing; the method comprises the following steps:
the voice component weight and the video component weight of the user are independent, and the voice component weight and the video component weight are obtained separately;
sequentially calculating component weights for video data and voice data: respectively calculating covariance matrixes of video data and voice data of a user, and carrying out eigenvalue decomposition on the covariance matrixes to obtain n eigenvalues mu= { mu 1 ,μ 2 ,...,μ n Sum n eigenvectors v= { v 1 ,v 2 ,...,v n }；
Dividing each eigenvalue by the sum of all eigenvalues to obtain a corresponding variance contribution ratio sigma i The formula is as follows:
multiplying each variance contribution rate by all components of the corresponding feature vector yields a weighted feature vector v' as follows:
v′＝{v 1 ×σ 1 ,...,v n ×σ n }；
summing each row of the feature vector v' results in a set of weighted values w, the formula being as follows:
w＝{w 1 ,w 2 ,...,w n }；
and normalizing the weighted values to obtain component weight, wherein the formula is as follows:
weight＝{w 1 ′,w′ 2 ,...,w′ n }；
training an integrated self-encoder;
obtaining a reconstruction weight through a reconstruction error of the integrated self-encoder; the method comprises the following steps:
reservoir sampling is carried out on the user data, n pieces of user data are obtained through each sampling, and an n multiplied by f user data matrix A is obtained; wherein f represents the number of features;
transpose the user data matrix A to obtain an f×n matrix A T ；
The total is sampled m times to obtain a f multiplied by m multiplied by n user matrix B;
the method comprises the steps of importing a self-encoder in pre-training, calculating reconstruction errors of m samples for each index, and solving an average value as a single sampling reconstruction error of the index;
repeating the reconstruction error operation for h times by adopting a bagging concept of an integrated algorithm to obtain a matrix C of h multiplied by f, and averaging the columns of the matrix C to obtain a reconstruction weight;
obtaining a user perception score: the average value of the component weights and the reconstruction weights is calculated to be used as an index weight, a user score is obtained according to the data of each user and the data value of each index, and the user score of each index is multiplied by the corresponding index weight to obtain a user perception score; the method comprises the following steps:
for a forward index, i.e., an index whose score should be higher the larger the value, the formula is as follows:
wherein x is the data value of each index; a and B are the lowest threshold and the highest threshold of the score respectively; if the data value is lower than A, the index score is 0; if the data value is higher than B, the index score is 100; the data value is between A and B, thenPerforming score calculation;
for negative indicators, i.e. indicators where the score should be higher the smaller the value, the formula is as follows:
wherein x is the data value of each index; a and B are the lowest threshold and the highest threshold of the score respectively; if the data value is higher than A, the index score is 0; if the data value is lower than B, the index score is 100; the data value is between A and B, thenPerforming score calculation;
the user score of each index is multiplied by the corresponding index weight to obtain a user perception score.
2. The method for user perception assessment based on PCA and integrated self-encoder according to claim 1, wherein the video-related KQI metrics include uplink average RTTms, downlink average RTTms, streaming media XKB start delay, streaming media pause duty cycle, streaming media pause frequency tinsmin, uplink TCP disorder rate, downlink TCP disorder rate, uplink TCP retransmission rate, downlink TCP retransmission rate, and streaming media downlink rate kbps;
the voice related KQI index VOLTE initial call network call completing rate, VOLTE response call dropping rate, uplink average MOS, downlink average MOS, uplink RTP packet dropping rate, downlink RTP packet dropping rate, uplink single pass proportion, downlink single pass proportion, uplink word swallowing proportion, downlink word swallowing proportion, uplink intermittent proportion and downlink intermittent proportion.
3. The method for evaluating user perception based on PCA and integrated self-encoder according to claim 1, wherein the normalization process using z-score method is specifically as follows:
obtaining a mean value and a standard value;
the data for each index is subtracted from the corresponding mean value and divided by the standard value so that the data meets the standard normal distribution.
4. The method of user perception assessment based on PCA and integrated self-encoder according to claim 1, characterized in that the training integrated self-encoder is in particular as follows:
normal value selection: the self-encoder is used for detecting abnormal values, and training data which are all normal values are needed;
selecting a normal value by adopting a 3 sigma method;
calculating the mean value theta of the data of the ith index i And standard deviation sigma i Will have a value of θ i ±3×σ i The data within are identified as normal values;
randomly sampling all normal values with replacement, and obtaining n normal values each time to form a sample until a moderate-scale data set is obtained;
training a self-encoder by using the constructed normal value to reconstruct samples, and establishing a mapping between original samples and reconstructed samples of the normal value;
a self-encoder trained with normal values will have difficulty reconstructing the outlier samples, one for each index.
5. A PCA and integrated self-encoder based user perception assessment system for implementing the PCA and integrated self-encoder based user perception assessment method of any of claims 1-4; the system includes a first processor configured to receive a signal,
the index acquisition module is used for acquiring a KQI index related to the voice of the high-speed rail user and a KQI index related to the video through the sensor;
the data cleaning module is used for cleaning data and specifically comprises the following steps: when a null value exists under the VOLTE response call drop rate index of one piece of data, setting the null value to zero; when the other arbitrary indexes in one piece of data have blank values, deleting the piece of user data and not participating in modeling;
the pretreatment module is used for carrying out pretreatment on the cleaned data, and specifically comprises the following steps: performing standardization processing by using a z-score method, wherein standardized data are used for weight calculation;
the component weight acquisition module is used for acquiring initial perception by using PCA and acquiring component weights through normalization processing;
the training module is used for training the integrated self-encoder;
the reconstruction weight acquisition module is used for acquiring reconstruction weights through reconstruction errors of the integrated self-encoder;
the user perception score acquisition module is used for solving the mean value of the component weights and the reconstruction weights to serve as index weights, acquiring user scores according to the data of each user and the data value of each index, and multiplying the user scores of each index with the corresponding index weights to obtain the user perception scores.
6. An electronic device, comprising: a memory and at least one processor;
wherein the memory has a computer program stored thereon;
the at least one processor executing the computer program stored by the memory causes the at least one processor to perform the PCA and integrated self-encoder based user perception assessment method as in any of claims 1-4.
7. A computer readable storage medium having stored therein a computer program executable by a processor to implement the PCA and integrated self-encoder based user perception assessment method according to any of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211452533.4A CN115842812B (en) | 2022-11-21 | 2022-11-21 | User perception evaluation method and system based on PCA and integrated self-encoder |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211452533.4A CN115842812B (en) | 2022-11-21 | 2022-11-21 | User perception evaluation method and system based on PCA and integrated self-encoder |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115842812A CN115842812A (en) | 2023-03-24 |
| CN115842812B true CN115842812B (en) | 2024-04-12 |
Family
ID=85577210
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211452533.4A Active CN115842812B (en) | 2022-11-21 | 2022-11-21 | User perception evaluation method and system based on PCA and integrated self-encoder |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115842812B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109768888A (en) * | 2019-01-16 | 2019-05-17 | 广东工业大学 | A kind of network service quality evaluation method, device, equipment and readable storage medium storing program for executing |
| CN112203332A (en) * | 2020-09-27 | 2021-01-08 | 南京嘉环科技股份有限公司 | VoLTE accurate switching method and system based on user perception |
| WO2021218828A1 (en) * | 2020-04-27 | 2021-11-04 | 支付宝(杭州)信息技术有限公司 | Training for differential privacy-based anomaly detection model |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110365503B (en) * | 2018-03-26 | 2022-08-19 | 华为技术有限公司 | Index determination method and related equipment thereof |
| US11399202B2 (en) * | 2019-04-05 | 2022-07-26 | Q'ligent Corporation | Analytics in video/audio content distribution networks |
-
2022
- 2022-11-21 CN CN202211452533.4A patent/CN115842812B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109768888A (en) * | 2019-01-16 | 2019-05-17 | 广东工业大学 | A kind of network service quality evaluation method, device, equipment and readable storage medium storing program for executing |
| WO2021218828A1 (en) * | 2020-04-27 | 2021-11-04 | 支付宝(杭州)信息技术有限公司 | Training for differential privacy-based anomaly detection model |
| CN112203332A (en) * | 2020-09-27 | 2021-01-08 | 南京嘉环科技股份有限公司 | VoLTE accurate switching method and system based on user perception |
Non-Patent Citations (2)
| Title |
|---|
| 利用主成分与MOS法综合评定IPTV客户感知水平;张应福;冯杰;刘劲;;通信与信息技术;20130125(01);34-38 * |
| 基于众包感知的OTT业务体验影响因素分析;徐小龙;李克;王海;宋晓勤;王征;吴雄;;数据采集与处理;20180515(03);180-190 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115842812A (en) | 2023-03-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107943583B (en) | Application processing method and device, storage medium and electronic equipment | |
| US11875558B2 (en) | Methods and apparatus to generate temporal representations for action recognition systems | |
| US20150161518A1 (en) | System and Method for Non-Invasive Application Recognition | |
| CN113379071B (en) | Noise label correction method based on federal learning | |
| CN105869022B (en) | Application popularity prediction method and device | |
| CN111340233B (en) | Training method and device of machine learning model, and sample processing method and device | |
| CN107728772B (en) | Application processing method and device, storage medium and electronic equipment | |
| CN114528190B (en) | Single index abnormality detection method and device, electronic equipment and readable storage medium | |
| CN113094284A (en) | Application fault detection method and device | |
| WO2018192414A1 (en) | Qoe determining method and apparatus, storage medium, and processor | |
| Frazier-Logue et al. | Dropout is a special case of the stochastic delta rule: faster and more accurate deep learning | |
| CN115842812B (en) | User perception evaluation method and system based on PCA and integrated self-encoder | |
| CN108765010B (en) | User retention probability prediction method and device, prediction server and storage medium | |
| CN111368858B (en) | User satisfaction evaluation method and device | |
| CN114827951A (en) | Vehicle network quality analysis method and system based on vehicle terminal and storage medium | |
| EP3331198A1 (en) | Method and device for monitoring a dsl network | |
| US20180157623A1 (en) | Sensing data based estimation method and sensing data based estimation system | |
| CN114842382A (en) | Method, device, equipment and medium for generating semantic vector of video | |
| CN113420165A (en) | Training of two-classification model and classification method and device of multimedia data | |
| CN112597764A (en) | Text classification method and device, storage medium and electronic device | |
| CN112286758A (en) | Information processing method, information processing device, electronic equipment and computer readable storage medium | |
| Sandholm et al. | MASS: Mobile Autonomous Station Simulation | |
| US20220374780A1 (en) | Training device, training method, and training program | |
| CN112634946B (en) | Voice quality classification prediction method, computer equipment and storage medium | |
| CN111309706A (en) | Model training method and device, readable storage medium and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |