A personalized growth plan generation method based on intelligent multi-dimensional evaluation

Abstract

The present invention discloses a method for generating a personalized growth plan for intelligent multi-dimensional evaluation, including: collecting students' response text data and response EEG data for different interest fields and preprocessing them; extracting features from the preprocessed data to obtain response text feature vectors and response EEG feature vectors; performing multimodal feature fusion on the feature vectors to obtain fused feature vectors; constructing an excitement point recognition model to identify the excitement points in the interest field for the fused feature vectors; and selecting a growth plan that meets the interest field from a preset growth plan library as a personalized growth plan generation result. The present invention extracts features from multi-dimensional evaluation data based on self-attention weights and time-frequency attention weight sequences, uses a graph neural network to take different interest fields as nodes, calculates the attention of different nodes and converts them into feature information of the nodes, and obtains a personalized growth plan based on the probability distribution by calculating the probability distribution of the excitement points in the interest field.

Description

Technical Field

[0001] The present invention relates to the technical field of personalized multi-dimensional evaluation, and in particular to a method for generating a personalized growth plan for intelligent multi-dimensional evaluation. Background Art

[0002] Faced with a new round of educational reforms, the traditional one-size-fits-all, homogenized education evaluation model no longer meets the needs of modern talent training, and personalized education is leading the new direction of educational development. Personalized growth is an educational model that tailors exclusive learning plans for students based on their individual personality traits and learning needs, aiming to promote the all-round development of students, improve their interest and motivation in learning, and cultivate their independent learning and innovation abilities. Traditional educational methods are mainly academic evaluations based on paper-and-pencil tests, which focus on students' cognitive knowledge memory and understanding, and do not pay enough attention to students' higher-order cognitive abilities. It is impossible to measure students' non-cognitive abilities, such as teamwork, emotional regulation, open-mindedness, social interaction, and conscientious performance. Summary of the invention

[0003] In response to the above problems, the present invention proposes a personalized growth plan generation method based on intelligent multi-dimensional assessment. The purpose is to accurately characterize the learners' personalized characteristics by paying attention to the learners' ontological characteristics analysis, knowledge mastery basis analysis, cognitive ability process analysis, emotional state analysis and learning behavior analysis on the basis of technologies such as artificial intelligence and big data analysis, so as to provide a personalized basis for the specification of growth plans.

[0004] In order to achieve the above object, the present invention provides a method for generating a personalized growth plan based on intelligent multi-dimensional evaluation, comprising the following steps:

[0005] S1: Collect students’ response text data and response EEG data on different areas of interest and pre-process them;

[0006] S2: extracting features from the preprocessed response text data and response EEG data to obtain a response text feature vector and a response EEG feature vector;

[0007] S3: performing multimodal feature fusion on the response text feature vector and the response EEG feature vector to obtain a fused feature vector;

[0008] S4: Construct an excitement point recognition model to identify the excitement points in the field of interest based on the fused feature vector;

[0009] S5: According to the identified excitement points in the field of interest, a growth plan that matches the field of interest is selected from a preset growth plan library as a personalized growth plan generation result.

[0010] As a further improvement method of the present invention:

[0011] Optionally, the step S1 includes:

[0012] S11: Collect students’ response text data and response EEG data on different areas of interest:

[0013] Media data in different interest areas are played, and students write their thoughts based on the played media data as response text data, and collect students' response EEG data during the media data playback process. The response text data and response EEG data are expressed as:

[0014] ;

[0015] in: represents the student's response text data for the nth area of ​​interest, represents the EEG data of the student's response to the nth area of ​​interest, and N represents the total number of areas of interest;

[0016] S12: Preprocessing the collected response text data and response EEG data to obtain preprocessed response text data and response EEG data.

[0017] Optionally, the step S12 includes:

[0018] S121: Response text data Preprocessing:

[0019] Response text data Perform word segmentation and remove stop words from the word segmentation results to obtain response text data The word segmentation result after removing stop words is used as the corresponding response text data The preprocessing results are:

[0020] ;

[0021] in: Represents response text data The preprocessing results are Represents response text data The i-th word segmentation result after removing stop words, Represents response text data The number of word segmentation results after removing stop words, ;

[0022] S122: Response to EEG data Preprocessing:

[0023] Response to EEG data Discretize it:

[0024] ;

[0025] in: Represents the response EEG data The discretization processing result is to sample the response EEG data. Represents the response EEG data The signal values ​​corresponding to the M sampling moments after discretization processing, where Represents the response EEG data The signal value corresponding to the mth sampling moment is, ;

[0026] Calculated The amplitude spectrum of:

[0027] ;

[0028] in: express The amplitude spectrum of represents the time interval between adjacent sampling moments, Represents the response EEG data The frequency of j represents the imaginary unit, ;

[0029] Using autoencoder network to respond to EEG data Perform self-encoding processing to obtain encoding results at different scales and respond to EEG data The autoencoding processing formula at the decomposition scale d is:

[0030] ;

[0031] in: Represents the response EEG data The result of the autoencoding process at the decomposition scale d is: , D represents the preset maximum decomposition scale, represents the autoencoder matrix, represents the bias parameter, T represents the transpose, represents the convolution process, Represents the activation function, and the selected activation function is the ReLU function;

[0032] The dynamic weights of the autoencoding processing results at different decomposition scales are calculated, where the autoencoding processing results The dynamic weight is:

[0033] ;

[0034] in: represents the L1 norm, Indicates the result of self-encoding processing Dynamic weight of

[0035] Based on the dynamic weights, different autoencoding processing results are weighted to obtain the preprocessed response EEG data:

[0036] ;

[0037] in: Represents the response EEG data after preprocessing.

[0038] Optionally, step S2 includes:

[0039] S21: Response text data after preprocessing Perform feature extraction:

[0040] Preprocessed response text data Each word segmentation result after removing stop words in is represented by word vector encoding to form an encoding vector:

[0041] ;

[0042] in: Represents the preprocessed response text data The corresponding encoding vector, The word vector encoding represents the result of the i-th word segmentation result after removing stop words, where the selected word vector model is the word2vec model;

[0043] The encoding vector Convert to matrix form :

[0044] ;

[0045] Among them: Matrix for OK A matrix of columns, Represents the element value of the i-th row and i-th column of the matrix;

[0046] Based on the encoding vector The corresponding matrix form is used to calculate the word vector encoding representation result The self-attention weights are:

[0047] ;

[0048] in: Represents element value The Euclidean distance between represents an exponential function with a natural constant as base, Represents the word vector encoding result The self-attention weight of

[0049] Preprocessed response text data based on self-attention weights The corresponding response text feature vector:

[0050] ;

[0051] in: Represents the preprocessed response text data The corresponding response text feature vector.

[0052] Optionally, the step S2 further includes:

[0053] S22: Response EEG data after preprocessing Perform feature extraction:

[0054] The preprocessed response EEG data Input into the LSTM model, use the LSTM model to calculate the hidden state of the signal value corresponding to different sampling times, and use Represents the preprocessed response EEG data The hidden state of the signal value corresponding to the mth sampling moment in ;

[0055] The preprocessed response EEG data is obtained based on the hidden state calculation The time-frequency attention weight sequence of:

[0056] ;

[0057] ;

[0058] ;

[0059] in: Represents the preprocessed response EEG data The time-frequency attention weight sequence, Represents the preprocessed response EEG data The time-frequency attention at the mth sampling moment in , Indicates hidden state With the final hidden state The similarity between represents the L1 norm;

[0060] Response EEG data after preprocessing Perform time-frequency attention weighting and perform residual convolution processing on the time-frequency attention weighting results to obtain the response EEG feature vector:

[0061] ;

[0062] in: Represents the preprocessed response EEG data The time-frequency attention weighted result of represents the convolution operation, Represents the preprocessed response EEG data The corresponding response EEG feature vector.

[0063] Optionally, step S3 includes:

[0064] S31: Compress and concatenate the response text feature vectors and the response EEG feature vectors corresponding to the N interest areas to obtain the compressed feature vectors corresponding to the N interest areas, where the response text feature vector corresponding to the nth interest area is and the response EEG feature vector The compression splicing formula is: ;

[0065] in: represents the compressed feature vector of the nth area of ​​interest, Represents the response text feature vector The linear compression result is: represents the text feature transformation matrix, Represents the text feature transformation bias, Represents the response EEG feature vector The linear compression result is: represents the EEG feature transformation matrix, represents the EEG feature transformation bias, Represents a splicing operation;

[0066] S32: Activate the compressed feature vector, where the compressed feature vector The activation processing formula is:

[0067] ;

[0068] in: represents the simplified feature transformation matrix, Represents the compressed feature vector The activation weight of

[0069] S33: Constructing fusion feature vector :

[0070] ;

[0071] ;

[0072] in: Represents the vector activation fusion result of the response text feature vector and the response EEG feature vector corresponding to the nth area of ​​interest.

[0073] Optionally, step S4 includes:

[0074] S41: Constructing an excitement point recognition model:

[0075] The excitement point recognition model takes the fusion feature vector as input and the probability distribution of the excitement points in the field of interest as output. The excitement point recognition model includes an input layer, a graph neural modeling layer, a probability distribution calculation layer of the excitement points in the field of interest, and an output layer. The probability distribution calculation layer of the excitement points in the field of interest includes an attention layer and an association matching layer.

[0076] S42: Fusion feature vector Identify the excitement points in the field of interest:

[0077] S421: The input layer receives the fused feature vector ;

[0078] S422: The graph neural modeling layer uses N interest areas as nodes and the vector activation fusion results corresponding to the N interest areas as node features to construct a graph neural network:

[0079] ;

[0080] Where: E represents the node set, , represents the nth node in the node set E, corresponding to the nth field of interest, V represents the node feature set, , Representation Node The node characteristics of ;

[0081] S423: The association matching layer calculates the importance between different nodes and provides the importance between nodes to the attention layer, where the nodes With Node The importance between them is expressed as:

[0082] ;

[0083] in: Representation Node With Node The importance of , represents the weight coefficient matrix;

[0084] The attention layer calculates the attention weights of different nodes based on the attention mechanism, and constructs the probability expressions of different nodes based on the Softmax function as the probability distribution of the excitement points in the field of interest. The attention weight is expressed as:

[0085] ;

[0086] ;

[0087] in: Representation Node The attention weight, Representation Node With Node Attention association weights between them;

[0088] The probability distribution of the excitement points in the field of interest is expressed as:

[0089] ;

[0090] ;

[0091] in: represents the probability distribution of excitement points in the field of interest, represents the probability of excitement point in the nth area of ​​interest, Represents the attention weight corresponding to the nth area of ​​interest characteristic information;

[0092] S424: The output layer outputs the probability distribution of the excitement points in the field of interest As the excitement point of the interest area corresponding to the fused feature vector F.

[0093] Optionally, step S5 includes:

[0094] S51: Obtain the distribution of importance weight coefficients of each growth plan in the growth plan library at different points of interest, wherein the distribution of importance weight coefficients of the u-th growth plan is:

[0095] ;

[0096] in: It represents the distribution of the weight coefficient of the u-th growth plan, represents the proportion of content associated with the nth interest point in the uth growth plan;

[0097] S52: Calculate the similarity between the importance weight coefficient distribution corresponding to each growth plan and the excitement point P in the interest field, where the importance weight coefficient distribution corresponding to the u-th growth plan The similarity with the excitement point P in the field of interest is:

[0098] ;

[0099] in: Indicates that the weight coefficient distribution is important Similarity with the excitement point P in the field of interest;

[0100] S53: Selecting the growth plan corresponding to the distribution of importance weight coefficients with the highest similarity as the growth plan that meets the interest area as the personalized growth plan generation result.

[0101] The present invention provides an electronic device, comprising:

[0102] A memory storing at least one instruction;

[0103] Communication interface, enabling electronic equipment to communicate; and

[0104] The processor executes the instructions stored in the memory to implement the personalized growth plan generation method of the above-mentioned intelligent multi-dimensional evaluation.

[0105] The present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the personalized growth plan generation method of the above-mentioned intelligent multi-dimensional assessment.

[0106] Compared with the prior art, the technical solution of the present invention has at least the following advantages:

[0107] 1. The present invention proposes a dynamically weighted autoencoding processing method, which adopts an autoencoding network to perform autoencoding processing on the response EEG data to obtain encoding results at different scales, and uses the ratio between the norm information of the encoding result and the amplitude spectrum as the dynamic weight. The larger the dynamic weight, the more amplitude information of the response EEG data contained in the encoding result, and the dynamic weighted values ​​of the encoding results at different scales are obtained. Based on the self-attention weight of the preprocessed response text data and the time-frequency attention weight sequence of the preprocessed response EEG data, feature extraction based on the attention mechanism is performed on the response text data and the response EEG data, so as to increase the attention to the key evaluation results and improve the evaluation accuracy.

[0108] 2. The present invention proposes a method of using an excitement point recognition model to identify excitement points in interest fields using a fused feature vector. First, linear compression is performed on the response text feature vector and the response EEG feature vector, and activation weights are calculated to obtain vector activation fusion results for different interest fields to form a fused feature vector. Then, a graph neural network is used to take different interest fields as nodes, and the graph attention of different nodes is calculated. The graph attention is converted into feature information of the nodes to realize the probability distribution calculation of excitement points in interest fields. Finally, the growth plan with the highest similarity to the probability distribution of excitement points in interest fields is selected as the personalized growth plan generation result, thereby realizing the generation of personalized growth plans under text response evaluation and EEG response evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] Figure 1 This is a flow chart of a method for generating a personalized growth plan based on intelligent multi-dimensional assessment according to Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0110] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0111] The execution subject of the personalized growth plan generation method of the intelligent multi-dimensional evaluation includes but is not limited to at least one of the electronic devices such as the server and the terminal that can be configured to execute the method provided in the embodiment of the present application. In other words, the personalized growth plan generation method of the intelligent multi-dimensional evaluation can be executed by software or hardware installed on the terminal device or the server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0112] Embodiment 1:

[0113] This embodiment provides a method for generating a personalized growth plan based on intelligent multi-dimensional assessment, including the following steps:

[0114] S1: Collect students' response text data and response EEG data in different areas of interest and pre-process them:

[0115] S11: Collect students’ response text data and response EEG data on different areas of interest:

[0116] Media data in different interest areas are played, and students write their thoughts based on the played media data as response text data, and collect students' response EEG data during the media data playback process. The response text data and response EEG data are expressed as:

[0117] ;

[0118] in: represents the student's response text data for the nth area of ​​interest, represents the EEG data of the student's response to the nth area of ​​interest, and N represents the total number of areas of interest;

[0119] S12: Preprocessing the collected response text data and response EEG data to obtain preprocessed response text data and response EEG data.

[0120] S121: Response text data Preprocessing:

[0121] Use word segmentation tools to respond to text data Perform word segmentation and remove stop words from the word segmentation results to obtain response text data The word segmentation result after removing stop words is used as the corresponding response text data The preprocessing results are:

[0122] ;

[0123] in: Represents response text data The preprocessing results are Represents response text data The i-th word segmentation result after removing stop words, Represents response text data The number of word segmentation results after removing stop words, ;

[0124] S122: Response to EEG data Preprocessing:

[0125] The main implementation method of EEG data preprocessing is based on a dynamic weighted autoencoder network;

[0126] Response to EEG data Discretize it:

[0127] ;

[0128] in: Represents the response EEG data The discretization processing result is to sample the response EEG data. Represents the response EEG data The signal values ​​corresponding to the M sampling moments after discretization processing, where Represents the response EEG data The signal value corresponding to the mth sampling moment is, ;

[0129] Calculated The amplitude spectrum of:

[0130] ;

[0131] in: express The amplitude spectrum of represents the time interval between adjacent sampling moments, Represents the response EEG data The frequency of j represents the imaginary unit, ;

[0132] Using autoencoder network to respond to EEG data Perform self-encoding processing to obtain encoding results at different scales and respond to EEG data The autoencoding processing formula at the decomposition scale d is:

[0133] ;

[0134] in: Represents the response EEG data The result of the autoencoding process at the decomposition scale d is: , D represents the preset maximum decomposition scale, represents the autoencoder matrix, represents the bias parameter, T represents the transpose, represents the convolution process, Represents the activation function, and the selected activation function is the ReLU function;

[0135] In this embodiment, ;

[0136] The dynamic weights of the autoencoding processing results at different decomposition scales are calculated, where the autoencoding processing results The dynamic weight is:

[0137] ;

[0138] in: represents the L1 norm, Indicates the result of self-encoding processing Dynamic weight of

[0139] Based on the dynamic weights, different autoencoding processing results are weighted to obtain the preprocessed response EEG data:

[0140] ;

[0141] in: Represents the response EEG data after preprocessing.

[0142] S2: Extract features from the preprocessed response text data and response EEG data to obtain response text feature vectors and response EEG feature vectors:

[0143] S21: Response text data after preprocessing Perform feature extraction:

[0144] Response text data after preprocessing Each word segmentation result after removing stop words in is represented by word vector encoding to form an encoding vector:

[0145] ;

[0146] in: Represents the preprocessed response text data The corresponding encoding vector, The word vector encoding represents the result of the i-th word segmentation result after removing stop words, where the selected word vector model is the word2vec model;

[0147] The encoding vector Convert to matrix form :

[0148] ;

[0149] Among them: Matrix for OK A matrix of columns, Represents the element value of the i-th row and i-th column of the matrix;

[0150] Based on the encoding vector The corresponding matrix form is used to calculate the word vector encoding representation result The self-attention weights are:

[0151] ;

[0152] in: Represents element value The Euclidean distance between represents an exponential function with a natural constant as base, Represents the word vector encoding result The self-attention weight of

[0153] Preprocessed response text data based on self-attention weights The corresponding response text feature vector:

[0154] ;

[0155] in: Represents the preprocessed response text data The corresponding response text feature vector.

[0156] S22: Response EEG data after preprocessing Perform feature extraction:

[0157] The main implementation method for feature extraction of EEG data is to use multi-scale and multi-path feature enhancement based on time-frequency attention mechanism;

[0158] The preprocessed response EEG data Input into the LSTM model, use the LSTM model to calculate the hidden state of the signal value corresponding to different sampling times, and use Represents the preprocessed response EEG data The hidden state of the signal value corresponding to the mth sampling moment in ;

[0159] The preprocessed response EEG data is obtained based on the hidden state calculation The time-frequency attention weight sequence of:

[0160] ;

[0161] ;

[0162] ;

[0163] in: Represents the preprocessed response EEG data The time-frequency attention weight sequence, Represents the preprocessed response EEG data The time-frequency attention at the mth sampling moment in , Indicates hidden state With the final hidden state The similarity between represents the L1 norm;

[0164] Response EEG data after preprocessing Perform time-frequency attention weighting and perform residual convolution processing on the time-frequency attention weighting results to obtain the response EEG feature vector:

[0165] ;

[0166] in: Represents the preprocessed response EEG data The time-frequency attention weighted result of represents the convolution operation, Represents the preprocessed response EEG data The corresponding response EEG feature vector.

[0167] S3: Perform multimodal feature fusion on the response text feature vector and the response EEG feature vector to obtain a fused feature vector:

[0168] S31: Compress and concatenate the response text feature vectors and the response EEG feature vectors corresponding to the N interest areas to obtain the compressed feature vectors corresponding to the N interest areas, where the response text feature vector corresponding to the nth interest area is and the response EEG feature vector The compression splicing formula is: ;

[0169] in: represents the compressed feature vector of the nth area of ​​interest, Represents the response text feature vector The linear compression result is: represents the text feature transformation matrix, Represents the text feature transformation bias, Represents the response EEG feature vector The linear compression result is: represents the EEG feature transformation matrix, represents the EEG feature transformation bias, Represents a splicing operation;

[0170] S32: Activate the compressed feature vector, where the compressed feature vector The activation processing formula is:

[0171] ;

[0172] in: represents the simplified feature transformation matrix, Represents the compressed feature vector The activation weight of

[0173] S33: Constructing fusion feature vector :

[0174] ;

[0175] ;

[0176] in: Represents the vector activation fusion result of the response text feature vector and the response EEG feature vector corresponding to the nth area of ​​interest.

[0177] S4: Construct an excitement point recognition model to identify the excitement points in the field of interest based on the fused feature vector:

[0178] S41: Constructing an excitement point recognition model:

[0179] The excitement point recognition model takes the fusion feature vector as input, the graph neural network as the main implementation method, and the probability distribution of the excitement points in the field of interest as output. The excitement point recognition model includes an input layer, a graph neural modeling layer, a probability distribution calculation layer of the excitement points in the field of interest, and an output layer. The probability distribution calculation layer of the excitement points in the field of interest includes an attention layer and an association matching layer;

[0180] S42: Fusion feature vector Identify the excitement points in the field of interest:

[0181] S421: The input layer receives the fused feature vector F;

[0182] S422: The graph neural modeling layer uses N interest areas as nodes and the vector activation fusion results corresponding to the N interest areas as node features to construct a graph neural network:

[0183] ;

[0184] Where: E represents the node set, , represents the nth node in the node set E, corresponding to the nth field of interest, V represents the node feature set, , Representation Node The node characteristics of ;

[0185] S423: The association matching layer calculates the importance between different nodes and provides the importance between nodes to the attention layer, where the nodes With Node The importance between them is expressed as:

[0186] ;

[0187] in: Representation Node With Node The importance of , represents the weight coefficient matrix;

[0188] The attention layer calculates the attention weights of different nodes based on the attention mechanism, and constructs the probability expressions of different nodes based on the Softmax function as the probability distribution of the excitement points in the field of interest. The attention weight is expressed as:

[0189] ;

[0190] ;

[0191] in: Representation Node The attention weight, Representation Node With Node Attention association weights between them;

[0192] The probability distribution of the excitement points in the field of interest is expressed as:

[0193] ;

[0194] ;

[0195] in: represents the probability distribution of excitement points in the field of interest, represents the probability of excitement point in the nth area of ​​interest, Represents the attention weight corresponding to the nth area of ​​interest The higher the probability of the excitement point, the richer the students’ response information is on the interest area, and the higher their attention is on the interest area.

[0196] S424: The output layer outputs the probability distribution of the excitement points in the field of interest As the excitement point of the interest area corresponding to the fused feature vector F.

[0197] S5: According to the identified excitement points in the field of interest, a growth plan that matches the field of interest is selected from a preset growth plan library as a personalized growth plan generation result:

[0198] S51: Obtain the distribution of importance weight coefficients of each growth plan in the growth plan library at different points of interest, wherein the distribution of importance weight coefficients of the u-th growth plan is:

[0199] ;

[0200] in: represents the distribution of the weight coefficients of the u-th growth plan, represents the proportion of content associated with the nth interest point in the uth growth plan;

[0201] S52: Calculate the similarity between the importance weight coefficient distribution corresponding to each growth plan and the excitement point P in the interest field, where the importance weight coefficient distribution corresponding to the u-th growth plan The similarity with the excitement point P in the field of interest is:

[0202] ;

[0203] in: Indicates that the weight coefficient distribution is important Similarity with the excitement point P in the field of interest;

[0204] S53: Selecting the growth plan corresponding to the distribution of importance weight coefficients with the highest similarity as the growth plan that meets the interest area as the personalized growth plan generation result.

[0205] Example 2

[0206] This embodiment provides an electronic device, which includes: a memory, storing at least one instruction; a communication interface, realizing electronic device communication; and a processor, executing the instructions stored in the memory to realize the personalized growth plan generation method of intelligent multi-dimensional assessment in Example 1.

[0207] The electronic device may include a processor, a memory, a communication interface and a bus, and may also include a computer program stored in the memory and executable on the processor.

[0208] The memory includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (for example, an SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. In some embodiments, the memory may be an internal storage unit of an electronic device, such as a mobile hard disk of the electronic device. In other embodiments, the memory may also be an external storage device of an electronic device, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory may also include both an internal storage unit of the electronic device and an external storage device. The memory may be used not only to store application software and various types of data installed in the electronic device, such as the code of a computer program, etc., but also to temporarily store data that has been output or is to be output.

[0209] In some embodiments, the processor may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips, etc. The processor is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect the various components of the entire electronic device, and executes or executes the programs or modules stored in the memory (computer programs for generating personalized growth plans for intelligent multi-dimensional evaluation, etc.), and calls the data stored in the memory to execute various functions of the electronic device and process data.

[0210] The communication interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices, and to achieve connection and communication between internal components of the electronic device.

[0211] The bus may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory and at least one processor, etc.

[0212] The electronic device may also include a power source (such as a battery) for supplying power to each component. Optionally, the power source may be logically connected to the at least one processor through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, etc. The electronic device may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0213] Optionally, the electronic device may further include a user interface, which may be a display (Display), an input unit (such as a keyboard (Keyboard)). Optionally, the user interface may also be a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device and to display a visual user interface.

[0214] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0215] The computer program stored in the memory of the electronic device is a combination of multiple instructions, and when executed in the processor, it can achieve:

[0216] Collect students' response text data and response EEG data in different areas of interest and pre-process them;

[0217] Perform feature extraction on the preprocessed response text data and response EEG data respectively to obtain a response text feature vector and a response EEG feature vector;

[0218] Perform multimodal feature fusion on the response text feature vector and the response EEG feature vector to obtain a fused feature vector;

[0219] Construct an excitement point recognition model to identify the excitement points in the field of interest based on the fused feature vector;

[0220] According to the identified excitement points in the field of interest, a growth plan that matches the field of interest is selected from a preset growth plan library as a personalized growth plan generation result.

[0221] Specifically, the specific implementation method of the processor for the above instructions can refer to Figure 1 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0222] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0223] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0224] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for generating a personalized growth plan based on intelligent multi-dimensional assessment, characterized in that: The method comprises the following steps: S1: Collect students’ response text data and response EEG data on different areas of interest and pre-process them; S2: extracting features from the preprocessed response text data and response EEG data to obtain a response text feature vector and a response EEG feature vector; S3: performing multimodal feature fusion on the response text feature vector and the response EEG feature vector to obtain a fused feature vector; S4: Construct an excitement point recognition model to identify the excitement points in the field of interest based on the fused feature vector; S5: According to the identified excitement points in the field of interest, a growth plan that matches the field of interest is selected from a preset growth plan library as a personalized growth plan generation result; The step S1 comprises: S11: Collect students’ response text data and response EEG data on different areas of interest: Media data in different interest areas are played, and students write their thoughts based on the played media data as response text data, and collect students' response EEG data during the media data playback process. The response text data and response EEG data are expressed as: ； in: represents the student's response text data for the nth area of ​​interest, represents the EEG data of the student's response to the nth area of ​​interest, and N represents the total number of areas of interest; S12: Preprocessing the collected response text data and response EEG data to obtain preprocessed response text data and response EEG data: S121: Response text data Preprocessing: Response text data Perform word segmentation and remove stop words from the word segmentation results to obtain response text data The word segmentation result after removing stop words is used as the corresponding response text data The preprocessing results are: ； in: Represents response text data The preprocessing results are Represents response text data The i-th word segmentation result after removing stop words, Represents response text data The number of word segmentation results after removing stop words, ; S122: Response to EEG data Preprocessing: Response to EEG data Discretize it: ； in: Represents the response EEG data The discretization processing result is to sample the response EEG data. Represents the response EEG data The signal values ​​corresponding to the M sampling moments after discretization processing, where Represents the response EEG data The signal value corresponding to the mth sampling moment is, ; Calculated The amplitude spectrum of: ； in: express The amplitude spectrum of represents the time interval between adjacent sampling moments, Represents the response EEG data The frequency of j represents the imaginary unit, ; Using autoencoder network to respond to EEG data Perform self-encoding processing to obtain encoding results at different scales and respond to EEG data The autoencoding processing formula at the decomposition scale d is: ； in: Represents the response EEG data The result of the autoencoding process at the decomposition scale d is: , D represents the preset maximum decomposition scale, represents the autoencoder matrix, represents the bias parameter, T represents the transpose, represents the convolution process, Represents the activation function, and the selected activation function is the ReLU function; The dynamic weights of the autoencoding processing results at different decomposition scales are calculated, where the autoencoding processing results The dynamic weight is: ； in: represents the L1 norm, Indicates the result of self-encoding processing Dynamic weight of Based on the dynamic weights, different autoencoding processing results are weighted to obtain the preprocessed response EEG data: ； in: Represents the preprocessed response EEG data.

2. The method for generating a personalized growth plan based on intelligent multi-dimensional assessment according to claim 1, characterized in that: The step S2 comprises: S21: Response text data after preprocessing Perform feature extraction: Response text data after preprocessing Each word segmentation result after removing stop words in is represented by word vector encoding to form an encoding vector: ； in: Represents the preprocessed response text data The corresponding encoding vector, The word vector encoding represents the result of the i-th word segmentation result after removing stop words, where the selected word vector model is the word2vec model; The encoding vector Convert to matrix form : ； Among them: Matrix for OK A matrix of columns, Represents the element value of the i-th row and i-th column of the matrix; Based on the encoding vector The corresponding matrix form is used to calculate the word vector encoding representation result The self-attention weights are: ； in: Represents element value The Euclidean distance between represents an exponential function with a natural constant as base, Represents the word vector encoding result The self-attention weight of Preprocessed response text data based on self-attention weights The corresponding response text feature vector: ； in: Represents the preprocessed response text data The corresponding response text feature vector.

3. The method for generating a personalized growth plan based on intelligent multi-dimensional assessment as claimed in claim 2, characterized in that: The step S2 further comprises: S22: Response EEG data after preprocessing Perform feature extraction: The preprocessed response EEG data Input into the LSTM model, use the LSTM model to calculate the hidden state of the signal value corresponding to different sampling times, and use Represents the preprocessed response EEG data The hidden state of the signal value corresponding to the mth sampling moment in ; The preprocessed response EEG data is obtained based on the hidden state calculation The time-frequency attention weight sequence of: ； ； ； in: Represents the preprocessed response EEG data The time-frequency attention weight sequence, Represents the preprocessed response EEG data The time-frequency attention at the mth sampling moment in , Indicates hidden state With the final hidden state The similarity between represents the L1 norm; Response EEG data after preprocessing Perform time-frequency attention weighting and perform residual convolution processing on the time-frequency attention weighting results to obtain the response EEG feature vector: ； in: Represents the preprocessed response EEG data The time-frequency attention weighted result of represents the convolution operation, Represents the preprocessed response EEG data The corresponding response EEG feature vector.

4. The method for generating a personalized growth plan based on intelligent multi-dimensional assessment as claimed in claim 3, characterized in that: The step S3 comprises: S31: Compress and concatenate the response text feature vectors and the response EEG feature vectors corresponding to the N interest areas to obtain the compressed feature vectors corresponding to the N interest areas, where the response text feature vector corresponding to the nth interest area is and the response EEG feature vector The compression splicing formula is: ; in: represents the compressed feature vector of the nth area of ​​interest, Represents the response text feature vector The linear compression result is: represents the text feature transformation matrix, Represents the text feature transformation bias, Represents the response EEG feature vector The linear compression result is: represents the EEG feature transformation matrix, represents the EEG feature transformation bias, Represents a splicing operation; S32: Activate the compressed feature vector, where the compressed feature vector The activation processing formula is: ； in: represents the simplified feature transformation matrix, Represents the compressed feature vector The activation weight of S33: Constructing fusion feature vector : ； ； in: Represents the vector activation fusion result of the response text feature vector and the response EEG feature vector corresponding to the nth area of ​​interest.

5. The method for generating a personalized growth plan based on intelligent multi-dimensional assessment as claimed in claim 4, characterized in that: The step S4 comprises: S41: Constructing an excitement point recognition model: The excitement point recognition model takes the fusion feature vector as input and the probability distribution of the excitement points in the field of interest as output. The excitement point recognition model includes an input layer, a graph neural modeling layer, a probability distribution calculation layer of the excitement points in the field of interest, and an output layer. The probability distribution calculation layer of the excitement points in the field of interest includes an attention layer and an association matching layer. S42: Fusion feature vector Identify the excitement points in the field of interest: S421: The input layer receives the fused feature vector ; S422: The graph neural modeling layer uses N interest areas as nodes and the vector activation fusion results corresponding to the N interest areas as node features to construct a graph neural network: ； Where: E represents the node set, , represents the nth node in the node set E, corresponding to the nth field of interest, V represents the node feature set, , Representation Node The node characteristics of ; S423: The association matching layer calculates the importance between different nodes and provides the importance between nodes to the attention layer, where the nodes With Node The importance between them is expressed as: ； in: Representation Node With Node The importance of , represents the weight coefficient matrix; The attention layer calculates the attention weights of different nodes based on the attention mechanism, and constructs the probability expressions of different nodes based on the Softmax function as the probability distribution of the excitement points in the field of interest. The attention weight is expressed as: ； ； in: Representation Node The attention weight, Representation Node With Node Attention association weights between them; The probability distribution of the excitement points in the field of interest is expressed as: ； ； in: represents the probability distribution of excitement points in the field of interest, represents the probability of excitement point in the nth area of ​​interest, Represents the attention weight corresponding to the nth area of ​​interest characteristic information; S424: The output layer outputs the probability distribution of the excitement points in the field of interest As the fusion feature vector Corresponding excitement points in areas of interest.

6. The method for generating a personalized growth plan based on intelligent multi-dimensional assessment as claimed in claim 5, characterized in that: The step S5 comprises: S51: Obtain the distribution of importance weight coefficients of each growth plan in the growth plan library at different points of interest, wherein the distribution of importance weight coefficients of the u-th growth plan is: ； in: It represents the distribution of the weight coefficient of the u-th growth plan, represents the proportion of content associated with the nth interest point in the uth growth plan; S52: Calculate the similarity between the importance weight coefficient distribution corresponding to each growth plan and the excitement point P in the interest field, where the importance weight coefficient distribution corresponding to the u-th growth plan The similarity with the excitement point P in the field of interest is: ； in: Indicates that the weight coefficient distribution is important Similarity with the excitement point P in the field of interest; S53: Selecting the growth plan corresponding to the distribution of importance weight coefficients with the highest similarity as the growth plan that meets the interest area as the personalized growth plan generation result.

Images