Artificial intelligence management system and methods for cross-platform data exchange and sharing

Abstract

The present invention relates to the field of data sharing, and in particular to an artificial intelligence management system and method for cross-platform data exchange and sharing. A data acquisition module is used to obtain platform data information and platform exchange and sharing signals corresponding to each platform; a data encryption module is used to encrypt the first platform data information through an AES encryption algorithm; a model building module is used to establish a target Bi LSTM+CRF recurrent neural network model; an address recognition module is used to obtain platform exchange and sharing signals, and input them into the target Bi LSTM+CRF recurrent neural network model for identification to obtain the target platform transmission address; an exchange and sharing module is used to obtain encrypted platform data information and target platform transmission address; data transmission time can be greatly shortened, data delay and loss during transmission can be reduced, and the real-time and accuracy of data transmission can be significantly improved.

Description

Technical Field

[0001] The present invention relates to the field of data sharing, and in particular to an artificial intelligence management system and method for cross-platform data exchange and sharing. Background Art

[0002] In cutting-edge fields like smart cities and intelligent transportation, parking lot vehicle management platforms, as a crucial component of intelligent urban infrastructure, are benefiting from the continued growth of network bandwidth. However, while network technology provides a robust foundation for data flow in parking lot vehicle management platforms, existing technology systems still face numerous challenges when faced with complex and diverse real-world application scenarios. First, the timeliness of cross-platform data transmission has become a bottleneck hindering the platform's efficient operation. Data formats and communication protocols often differ between vendors and systems, requiring multiple conversions and adaptations during data transmission. This not only increases transmission time but can also compromise data integrity and accuracy due to data loss during conversion. Second, data diversity exacerbates the difficulty of data exchange and sharing. Within parking lot vehicle management platforms, data types include but are not limited to vehicle information, owner information, entry and exit records, payment data, and parking space occupancy status. Each type of data carries specific business logic and value. However, due to differences in data formats, storage methods, and security levels, cross-system and cross-platform data exchange often encounters obstacles, making seamless integration and efficient sharing difficult, thereby increasing the complexity and cost of data processing. Therefore, how to achieve seamless docking and efficient sharing of vehicles in parking lots is a technical problem that needs to be solved at this stage. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and to design an artificial intelligence management system and method for cross-platform data exchange and sharing.

[0004] The technical solution of the present invention to achieve the above-mentioned purpose is that, further, in the above-mentioned artificial intelligence management system for cross-platform data exchange and sharing, the artificial intelligence management system for data exchange and sharing includes a data acquisition module, a data processing module, a data encryption module, a model building module, an address recognition module, an exchange and sharing module, and a transmission feedback module:

[0005] Data acquisition module, used to obtain platform data information corresponding to each platform and platform exchange sharing signals;

[0006] A data processing module, configured to perform data preprocessing on the platform data information to obtain first platform data information;

[0007] a data encryption module, configured to obtain the first platform data information, and encrypt the first platform data information using an AES encryption algorithm to obtain encrypted platform data information;

[0008] A model building module is used to establish an initial BiLSTM+CRF recurrent neural network model based on the RNN recurrent neural network, and replace the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model with the LeakyReLu activation function to obtain a target BiLSTM+CRF recurrent neural network model;

[0009] An address recognition module is used to obtain the platform exchange sharing signal, input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for recognition, and obtain the target platform transmission address;

[0010] An exchange sharing module, configured to obtain the encrypted platform data information and the target platform transmission address, and transmit the encrypted platform data information to the target platform transmission address;

[0011] The transmission feedback module is used to obtain the platform exchange feedback result. If the platform exchange feedback result is that the encrypted platform data information transmission address is incorrect, the platform exchange sharing signal is re-input into the target BiLSTM+CRF recurrent neural network model for identification.

[0012] Furthermore, in the above-mentioned artificial intelligence management system applied to cross-platform data exchange and sharing, the data acquisition module includes a platform data acquisition unit and an exchange signal acquisition unit:

[0013] A platform data acquisition unit is used to obtain platform data information corresponding to each platform. The platform data information is the operational data in the parking lot vehicle management platform, including vehicle information, owner information, entry and exit records, payment data, and parking space occupancy status;

[0014] The exchange signal acquisition unit is used to obtain data acquisition instructions from each platform system and generate platform exchange sharing information based on the data acquisition instructions; the data acquisition instructions include voice input instructions, text input instructions, and vehicle image input instructions.

[0015] Furthermore, in the above-mentioned artificial intelligence management system applied to cross-platform data exchange and sharing, the data processing module includes a data screening unit, a vectorization unit, a normalization unit, and a data classification unit:

[0016] A data screening unit is used to obtain platform data information corresponding to each platform, and use feature engineering to screen the platform features of the platform data information to obtain a platform feature label library;

[0017] A vectorization unit, configured to perform vectorization processing on the text information of the platform feature tag library by one-hot encoding, convert the information into a digital vector using an encoding method, and obtain a platform text feature tag library;

[0018] a normalization unit, configured to perform normalization processing on the platform text feature label library to obtain a target platform text feature label library, wherein the target platform text feature label library conforms to a standard normal distribution;

[0019] The data classification unit is used to classify the target platform text feature tag library through the FCM fuzzy clustering algorithm to obtain the first platform data information.

[0020] Furthermore, in the above-mentioned artificial intelligence management system for cross-platform data exchange and sharing, the data encryption module includes an encryption matrix unit, a round key addition unit, a substitution unit, a row shift unit, and a column obfuscation unit:

[0021] The encryption matrix unit is used to ensure that the 16-byte plaintext, 16-byte ciphertext and wheel key in the AES encryption algorithm are all a 4x4 matrix;

[0022] A round key adding unit is used to perform an XOR operation on each first platform data information in the matrix and the secondary round key, and each subkey is generated by the key;

[0023] a replacement unit, configured to replace each piece of first platform data information with corresponding second platform data information using a lookup table through a nonlinear replacement function;

[0024] A row shift unit, configured to cyclically shift the horizontal columns of each second platform data information in the matrix;

[0025] The column obfuscation unit is used to mix the four second platform data information in each column by using linear transformation to obtain encrypted platform data information.

[0026] Furthermore, in the above-mentioned artificial intelligence management system for cross-platform data exchange and sharing, the model building module includes a model building unit, a configuration parameter unit, a filter pruning unit, a threshold activation unit, an adaptability unit, and a model iteration unit:

[0027] A model building unit is used to establish an initial BiLSTM+CRF recurrent neural network model based on the RNN recurrent neural network, and replace the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model with a LeakyReLu activation function;

[0028] The configuration parameter unit is used to automatically search for the configuration parameters of the initial BiLSTM+CRF recurrent neural network model based on the CondAFP framework and the design space exploration of the genetic algorithm;

[0029] The filter pruning unit is used in the CondAFP framework to perform conditional activation-based model optimization and filter pruning-based model optimization strategies according to the gene expression of each individual chromosome in the initial BiLSTM+CRF recurrent neural network model, thereby obtaining a compressed initial BiLSTM+CRF recurrent neural network model.

[0030] The threshold activation unit is used to activate according to the threshold corresponding to each stage in the gene using entropy perception, predict input samples of different difficulty levels through multiple stages with different computational complexity, and remove unimportant filters to compress the initial BiLSTM+CRF recurrent neural network model;

[0031] Adaptability unit, used in the CondAFP framework to evaluate the adaptability of the corresponding individuals based on the accuracy and FLOPS of the compressed multi-stage model and rank the individuals according to their adaptability;

[0032] The model iteration unit is used to update the chromosome set to obtain the next generation of individuals, and then obtain the target BiLSTM+CRF recurrent neural network model through iteration.

[0033] Furthermore, in the above-mentioned artificial intelligence management system applied to cross-platform data exchange and sharing, the address recognition module includes a signal acquisition unit, a signal recognition unit, and a target transmission unit:

[0034] a signal acquisition unit, configured to acquire the platform exchange sharing signal, wherein the platform exchange sharing signal includes at least a platform information acquisition signal and a platform information output signal;

[0035] A signal recognition unit, configured to input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for recognition;

[0036] The BiLSTM layer in the target BiLSTM+CRF recurrent neural network model includes at least three gating units, namely the forget gate, input gate, and output gate. The calculation formula is:

[0037] f t =σ(W f *[h t-1 , x t ]+b f )

[0038] i t =σ(W i *[h t-1 , x t ]+b i )

[0039]

[0040] O t =σ(W o *[h t-1 , x t ]+b o )

[0041] h t =O t ⊙tanh(C t )

[0042] Among them, f t 、i t , O t 、C t represents the forget gate, input gate, output gate and memory cell at time t; W f 、W i 、W o 、W c Represents the weight matrix corresponding to different control gates; b f 、b i 、b o 、b c represents the bias vector; Represents the intermediate state of the input, x t represents the input vector at time t, h t Represents the output result at time t, σ represents the LeakyReLu activation function, tanh represents the hyperbolic tangent activation function, and ⊙ represents the dot product operator;

[0043] The CRF layer in the target BiLSTM+CRF recurrent neural network model performs joint probability analysis on the label sequence by analyzing the relationship between adjacent labels. The analysis process is as follows: Assume that P is the output matrix of the BiLSTM layer, n is the number of exchanged shared signals, if X = {x1, x2, ..., x n} represents the input exchange sharing signal sequence, Y = {y1, y2, ..., yn} represents the corresponding output label sequence, then the score score (X, Y) of the input sequence X corresponding to the predicted sequence Y is calculated as:

[0044]

[0045] in, Indicates that the i-th exchange sharing signal is predicted as label y i The score, Represents the label y i-1 Move to label y i The score, The larger the value, the more label y i-1The greater the possibility, the conditional probability of predicting the label sequence Y is P(Y / X), and the calculation formula is:

[0046]

[0047] in, represents the true label sequence, Y X Represents all possible label sequences, score(X,Y) represents the correct score of path prediction, Indicates the score of a path among all paths;

[0048] The Viterbi algorithm is used to convert the label sequence Y with the highest score into * As the labeling result of the target BiLSTM+CRF recurrent neural network model, the calculation formula is:

[0049]

[0050] The target transmission unit is used to obtain a target platform transmission address, where the target platform transmission address includes at least one platform transmission address or more than two platform transmission addresses.

[0051] Furthermore, in the above-mentioned artificial intelligence management system applied to cross-platform data exchange and sharing, the exchange and sharing module includes a data transmission unit, a platform receiving unit, a platform sharing unit, and a platform exchange unit:

[0052] A data transmission unit, configured to obtain the encrypted platform data information and the target platform transmission address, and transmit the encrypted platform data information to the target platform transmission address;

[0053] The platform receiving unit is configured to receive at least one encrypted platform data message, and if more than two encrypted platform data messages are received, transmit the encrypted platform data messages one by one on the platform;

[0054] The platform sharing unit is used to input the encrypted platform data into the shared database for storage. If the corresponding encrypted platform data information needs to be called, the encrypted platform data information is obtained through the shared database;

[0055] The platform exchange unit is used to obtain the exchange requirements of each platform and transmit the encrypted platform data information according to the exchange requirements.

[0056] Furthermore, in the above-mentioned artificial intelligence management system for cross-platform data exchange and sharing, the transmission feedback module includes a platform feedback unit, a data storage unit, and a data identification unit:

[0057] A platform feedback unit is used to obtain a platform exchange feedback result, if the platform exchange feedback result is an encrypted platform data information transmission address error;

[0058] A data storage unit is configured to transmit the encrypted platform data information to a shared database for storage if the platform exchange feedback result indicates that the transmission address of the encrypted platform data information is correct;

[0059] A data identification unit is used to re-input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for identification.

[0060] Furthermore, the artificial intelligence management method applied to cross-platform data exchange and sharing also includes the following steps:

[0061] Obtain platform data information corresponding to each platform and platform exchange and sharing signals;

[0062] performing data preprocessing on the platform data information to obtain first platform data information;

[0063] Acquire the first platform data information, and encrypt the first platform data information using an AES encryption algorithm to obtain encrypted platform data information;

[0064] An initial BiLSTM+CRF recurrent neural network model is established based on the RNN recurrent neural network, and the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model is replaced with the LeakyReLu activation function to obtain a target BiLSTM+CRF recurrent neural network model;

[0065] Obtain the platform exchange sharing signal, input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for recognition, and obtain the target platform transmission address;

[0066] Acquire the encrypted platform data information and the target platform transmission address, and transmit the encrypted platform data information to the target platform transmission address;

[0067] Obtain a platform exchange feedback result. If the platform exchange feedback result is that the encrypted platform data information transmission address is incorrect, re-input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for recognition.

[0068] Furthermore, the artificial intelligence management method applied to cross-platform data exchange and sharing also includes the following steps:

[0069] Obtain the platform data information corresponding to each platform, use feature engineering to filter the platform features of the platform data information, and obtain the platform feature label library;

[0070] Vectorizing the text information of the platform feature tag library through one-hot encoding, converting it into a digital vector using an encoding method, and obtaining a platform text feature tag library;

[0071] Normalizing the platform text feature label library to obtain a target platform text feature label library, wherein the target platform text feature label library conforms to a standard normal distribution;

[0072] The target platform text feature tag library is classified by using the FCM fuzzy clustering algorithm to obtain first platform data information.

[0073] Its beneficial effects are that, through the data acquisition module, it is used to obtain the platform data information and platform exchange sharing signals corresponding to each platform; the data processing module is used to perform data preprocessing on the platform data information to obtain the first platform data information; the data encryption module is used to obtain the first platform data information, encrypt the first platform data information through the AES encryption algorithm, and obtain the encrypted platform data information; the model building module is used to establish an initial BiLSTM+CRF recurrent neural network model based on the RNN recurrent neural network, and use the LeakyReLu activation function to replace the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model to obtain the target BiLSTM An M+CRF recurrent neural network model; an address recognition module for obtaining the platform exchange sharing signal and inputting it into the target BiLSTM+CRF recurrent neural network model for identification, thereby obtaining the target platform transmission address; an exchange sharing module for obtaining the encrypted platform data information and the target platform transmission address, and transmitting the encrypted platform data information to the target platform transmission address; and a transmission feedback module for obtaining the platform exchange feedback result. If the platform exchange feedback result indicates an error in the encrypted platform data information transmission address, the platform exchange sharing signal is re-inputted into the target BiLSTM+CRF recurrent neural network model for identification. 1. Significantly Improve Data Transmission Efficiency: By adopting standardized data exchange protocols and interface specifications, and optimizing cross-platform data transmission mechanisms, data transmission time can be significantly shortened, data delays and loss during transmission can be reduced, and the real-time and accuracy of data transmission can be significantly improved. 2. Reduced Data Exchange and Sharing Costs: Seamless data exchange and efficient sharing are achieved, reducing conversion and adaptation costs caused by data format mismatches and communication protocol differences. Furthermore, by utilizing advanced technologies such as cloud computing for data storage and processing, local storage and computing resource consumption can be effectively reduced, further reducing operating costs. 3. Improve urban management: As a crucial component of smart cities, the parking lot vehicle management platform's improved data transmission and sharing efficiency will help city managers better understand urban traffic dynamics, optimize traffic resource allocation, and improve urban traffic operation efficiency. At the same time, in-depth mining and analysis of massive parking data can provide strong support for urban planning and traffic policy formulation. 4. Strengthen data security and privacy protection: While improving data transmission efficiency and sharing capabilities, strengthening data security measures to ensure the security and privacy of data during transmission and sharing will help build user trust and protect user rights. This is of great significance for enhancing the reliability and credibility of the entire parking lot vehicle management platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Various other advantages and benefits will become apparent to those skilled in the art by reading the following detailed description of the preferred embodiment.The accompanying drawings are only for the purpose of illustrating the preferred embodiment and are not to be considered as limiting the present invention.

[0075] Figure 1 Schematic diagram of an embodiment of an artificial intelligence management system applied to cross-platform data exchange and sharing in an embodiment of the present invention;

[0076] Figure 2 Schematic diagram of a data encryption module according to an embodiment of the present invention;

[0077] Figure 3 This is a schematic diagram of an embodiment of a model building module in an embodiment of the present invention;

[0078] Figure 4 Schematic diagram of an embodiment of an artificial intelligence management method for cross-platform data exchange and sharing in an embodiment of the present invention. DETAILED DESCRIPTION

[0079] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0080] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a," "an," "said," and "said" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0081] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, the artificial intelligence management system for cross-platform data exchange and sharing includes the following modules:

[0082] Data acquisition module, used to obtain platform data information corresponding to each platform and platform exchange sharing signals;

[0083] Specifically, this embodiment includes a platform data acquisition unit for acquiring platform data information corresponding to each platform, where the platform data information is the operational data in the parking lot vehicle management platform; an exchange signal acquisition unit for acquiring data acquisition instructions in each platform system, and generating platform exchange sharing information based on the data acquisition instructions; data acquisition instructions include voice input instructions, text input instructions, and vehicle image input instructions.

[0084] A data processing module, configured to perform data preprocessing on the platform data information to obtain first platform data information;

[0085] Specifically, this embodiment includes a data screening unit, which is used to obtain platform data information corresponding to each platform, and use feature engineering to screen the platform features of the platform data information to obtain a platform feature label library; a vectorization unit, which is used to vectorize the text information of the platform feature label library through one-hot encoding, and convert it into a digital vector using an encoding method to obtain a platform text feature label library; a normalization unit, which is used to normalize the platform text feature label library to obtain a target platform text feature label library, and the target platform text feature label library conforms to the standard normal distribution; a data classification unit, which is used to classify the target platform text feature label library through the FCM fuzzy clustering algorithm to obtain the first platform data information.

[0086] Specifically, the Fuzzy C-Means (FCM) fuzzy clustering algorithm used in this embodiment is a clustering technique based on soft partitioning. It allows a data point to belong to multiple categories simultaneously with varying degrees of membership. This differs from traditional hard clustering algorithms (such as K-means), which require each data point to belong to a strict category. The FCM algorithm is suitable for processing data sets with fuzzy boundaries, meaning that a data point may not belong to any single category but rather be distributed across categories with a certain probability.

[0087] Basic Concepts

[0088] Membership: The core of the FCM algorithm lies in the membership assigned to each data point to each cluster center, which is a real number between 0 and 1 that indicates the degree to which the data point belongs to that category. The sum of the membership of each data point to all categories is usually equal to 1 (it may be different in some variants).

[0089] Objective function: The FCM algorithm determines the best cluster centers and membership matrix by minimizing an objective function (also called distortion function or sum of squared errors). This objective function measures the sum of the weighted distances from a data point to the center of each cluster to which it belongs.

[0090] Parameter m: An important parameter in the FCM algorithm, called the fuzzification parameter, controls the degree of fuzziness in the clustering. When m > 1, a larger m value will make the clustering results more fuzzy, that is, data points can more strongly belong to multiple categories; when it is close to 1, the clustering results tend to be hard clusters.

[0091] Iterative process: The FCM algorithm optimizes the clustering results by iterating two steps:

[0092] Update membership: For each data point, according to its distance to each cluster center and its current membership, recalculate its membership to each cluster center according to the formula.

[0093] Update the cluster center: Update the location of the cluster center based on the weighted average of all data points to the current cluster center (the weight is the respective membership).

[0094] Stopping condition: The iterative process usually stops when the preset maximum number of iterations is reached or the change in the objective function is less than a certain threshold.

[0095] A data encryption module is used to obtain the first platform data information and encrypt the first platform data information using an AES encryption algorithm to obtain encrypted platform data information;

[0096] Specifically, this embodiment includes an encryption matrix unit for determining that the 16-byte plaintext, 16-byte ciphertext, and round key in the AES encryption algorithm are all a 4x4 matrix; a round key addition unit for performing an XOR operation on each first platform data information in the matrix with the secondary round key, and each subkey is generated by the key; a substitution unit for replacing each first platform data information with the corresponding second platform data information using a lookup table through a nonlinear substitution function; a row shift unit for cyclically shifting the horizontal column of each second platform data information in the matrix; and a column obfuscation unit for mixing the four second platform data information in each column using a linear transformation to obtain encrypted platform data information.

[0097] Specifically, in this embodiment, AES (Advanced Encryption Standard) is a symmetric encryption algorithm used to protect the security of electronic data. It is an improvement on the Rijndael algorithm designed by Belgian cryptographers Vincent Rijmen and Joan Daemen. In 2001, it was selected by the U.S. National Institute of Standards and Technology (NIST) as a next-generation encryption algorithm to replace the original DES (Data Encryption Standard). AES has also been widely adopted internationally and has become the de facto standard in the field of data encryption.

[0098] Key Features: Symmetric Encryption: AES uses the same key for both encryption and decryption. Block Cipher: AES is a block cipher that encrypts plaintext into fixed-size blocks. AES supports a fixed block size of 128 bits (16 bytes). Key Length: AES supports three key lengths: 128 bits, 192 bits, and 256 bits, corresponding to AES-128, AES-192, and AES-256, respectively. Longer key lengths theoretically increase encryption strength, but also incur higher computational costs. Number of Encryption Rounds: AES uses different numbers of rounds to enhance security, depending on the key length. AES-128 uses 10 rounds, AES-192 uses 12 rounds, and AES-256 uses 14 rounds. Encryption Process: The AES encryption process consists of four main steps, which are performed in every round, except for the final round, which is slightly different. These four steps are: Byte Substitution (Subbytes): Each byte is nonlinearly permuted using S-boxes (substitution boxes). ShiftRows: Circularly shifts the rows of the state matrix. MixColumns: Mixes the columns of the state matrix using a linear transformation. AddRoundKey: XORs the round key with the state matrix. In the final encryption round, the column mix step is omitted, and only byte substitution and round key addition are performed.

[0099] Specifically, the XOR operation in this embodiment, that is, the exclusive OR operation, is a logical operation widely used in fields such as computer science, electronic engineering, and cryptography.

[0100] The model building module is used to build an initial BiLSTM+CRF recurrent neural network model based on the RNN recurrent neural network, and use the LeakyReLu activation function to replace the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model to obtain the target BiLSTM+CRF recurrent neural network model;

[0101] Specifically, the present embodiment includes a model building unit for building an initial BiLSTM+CRF recurrent neural network model based on the RNN recurrent neural network, and replacing the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model with the LeakyReLu activation function; a configuration parameter unit for automatically searching for the configuration parameters of the initial BiLSTM+CRF recurrent neural network model based on the CondAFP framework and the design space exploration of the genetic algorithm; a filter pruning unit for the CondAFP framework to respectively execute the model optimization strategy based on conditional activation and the filter pruning strategy based on the gene expression of each individual chromosome in the initial BiLSTM+CRF recurrent neural network model. The model optimization strategy of the branch is used to obtain a compressed initial BiLSTM+CRF recurrent neural network model; the threshold activation unit is used to activate according to the threshold corresponding to each stage in the gene using entropy perception, predict input samples of different difficulty levels through multiple stages with different computational complexities, and remove unimportant filters to compress the initial BiLSTM+CRF recurrent neural network model; the adaptability unit is used in the CondAFP framework to evaluate the adaptability of the corresponding individuals according to the accuracy and FLOPS of the compressed multi-stage model, and to sort the individuals according to their adaptability; the model iteration unit is used to update the chromosome set to obtain the next generation of individuals, and then obtain the target BiLSTM+CRF recurrent neural network model after iteration.

[0102] Specifically, in this embodiment, a BiLSTM+CRF recurrent neural network model is built based on PyTorh. The model is a dual-region LSTM neural network with a CRF layer. The Chinese character sequence is projected into a dense vector sequence with redundant features as the input of the recurrent layer. Here we use a hot vector representing the word boundary feature to illustrate. The recurrent layer is a bidirectional LSTM layer, and the outputs of the forward and backward vectors are summarized and projected as the score of each label. The CRF layer is used to overcome the label skew problem. The model is similar to the most advanced Chinese named entity recognition model in Chinese named entity recognition with basic level functions proposed in the character-based BLSTM-CRF. Among them, Bi.STM is an improved LSTM of the recurrent neural network RNN, which is particularly suitable for processing time-series text data. Relying on the super nonlinear fitting ability of the neural network and the transmission of information by the hidden state, LSTM transforms the samples through complex nonlinear transformations in high-dimensional space during training to learn the function from sample to annotation.

[0103] Specifically, the CondAFP framework in this embodiment is summarized as follows:

[0104] The CondAFP framework is an optimization strategy for deep learning models, particularly recurrent neural network (RNN) models, such as those combining bidirectional long short-term memory (BiLSTM) and conditional random fields (CRF). The framework's core concept is to incorporate biologically inspired ideas, mapping the principles of gene expression to the model optimization process. It improves model efficiency and performance through two mechanisms: conditional activation and filter pruning.

[0105] 1. Model Optimization Strategy Based on Conditional Activation: Inspired by Gene Expression: Drawing on the biological regulation of gene expression—the principle that genes are selectively expressed under different conditions—the CondAFP framework introduces the concept of "conditional activation." This may mean that in a BiLSTM model, the activation of certain neurons or layers is not constant but dynamically adjusted based on the specific conditions of the input data. For example, certain hidden units may only activate in specific contexts, reducing unnecessary computation and improving the model's relevance and efficiency.

[0106] Implementation: This can be achieved through a conditional gating mechanism (such as the update gate and reset gate in GRU, or the input gate, forget gate, and output gate in LSTM), which dynamically adjusts the information flow according to the input features to achieve conditional network activation.

[0107] 2. Model Optimization Strategy Based on Filter Pruning: Filter pruning is a model compression technique that reduces model complexity and computational requirements by removing weights or entire filters (in convolutional neural networks) that contribute little to the model output. While there are no "filters" in the traditional sense in the BiLSTM+CRF model, this can be understood as pruning neurons or hidden layers that contribute little to sequence feature recognition. Relationship with gene expression: This process can be likened to the natural selection and evolution of genes in organisms, removing redundant or unimportant gene expression while retaining gene functions that are critical for survival and adaptation to the environment.

[0108] Overall Framework Operation: The CondAFP framework combines these two optimization strategies to build a more efficient and adaptable model. Initially, a conditional activation strategy may be used to dynamically adjust the network structure to adapt to varying data characteristics. Subsequently, based on performance metrics and feedback from the model training process, filter pruning is used to remove redundant components, further streamlining the model and improving operational efficiency. This entire process emphasizes the model's adaptability and dynamics, mimicking the competitive and optimization mechanisms of nature to achieve improved generalization and resource efficiency.

[0109] Specifically, the LeakyReLU (Leaky Rectified Linear Unit) activation function in this embodiment is a variant of the ReLU (Rectified Linear Unit) activation function, which is mainly used to solve the problem that the gradient of the ReLU function is zero in the negative region, thereby alleviating the "neuron death" (Dead ReLU Problem) phenomenon in the neural network. In deep learning, activation functions are used to introduce nonlinearity to help neural networks learn more complex patterns. Advantages: Alleviating the gradient vanishing problem: Compared with ReLU, LeakyReLU has a non-zero gradient in the negative region, which can prevent neurons from "dying" during training, that is, the gradient is always zero, resulting in the inability to update the weights. High computational efficiency: Despite the introduction of a non-zero slope, LeakyReLU still maintains the high efficiency of linear operations and does not significantly increase the computational burden. Easy to implement: Like ReLU, the function form of LeakyReLU is simple and easy to understand and implement.

[0110] Specifically, in this embodiment, FLOPS (Floating Point Operations Per Second) refers to the number of floating-point operations per second, a key metric for measuring the computing power of a computer system or hardware device, particularly when performing high-performance computing tasks such as large-scale scientific computing, deep learning, and image processing. FLOPS assessment primarily focuses on the speed at which a machine performs floating-point operations, including addition, subtraction, multiplication, and division.

[0111] The initial BiLSTM+CRF recurrent neural network model has been trained using the platform exchange sharing signals that previously existed in the system, and is able to identify the platform exchange sharing signals and generate corresponding platform requirements and transmission addresses based on the platform exchange sharing signals.

[0112] The address recognition module is used to obtain the platform exchange sharing signal, input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for recognition, and obtain the target platform transmission address;

[0113] Specifically, this embodiment includes a signal acquisition unit for acquiring a platform exchange sharing signal, which includes at least a platform information acquisition signal and a platform information output signal; a signal recognition unit for inputting the platform exchange sharing signal into a target BiLSTM+CRF recurrent neural network model for recognition;

[0114] The BiLSTM layer in the target BiLSTM+CRF recurrent neural network model includes at least three gating units, namely the forget gate, input gate, and output gate. The calculation formula is:

[0115] ft =σ(W f *[h t-1 , x t ]+b f )

[0116] i t =σ(W i *[h t-1 , x t ]+b i )

[0117]

[0118] O t =σ(W o *[h t-1 , x t ]+b o )

[0119] h t =O t ⊙tanh(C t )

[0120] Among them, f t 、i t , O t 、C t represents the forget gate, input gate, output gate and memory cell at time t; W f 、W i 、W o 、W c Represents the weight matrix corresponding to different control gates; b f 、b i 、b o 、b c represents the bias vector; Represents the intermediate state of the input, x t represents the input vector at time t, h t Represents the output result at time t, σ represents the LeakyReLu activation function, tanh represents the hyperbolic tangent activation function, and ⊙ represents the dot product operator;

[0121] The CRF layer in the target BiLSTM+CRF recurrent neural network model performs joint probability analysis on the label sequence by analyzing the relationship between adjacent labels. The analysis process is as follows: Assume that P is the output matrix of the BiLSTM layer, n is the number of exchanged shared signals, if X = {x1, x2, ..., x n} represents the input exchange sharing signal sequence, Y={y1,y2,...,y n} represents the corresponding output label sequence, then the score (X, Y) of the input sequence X corresponding to the predicted sequence Y is calculated as:

[0122]

[0123] in, Indicates that the i-th exchange sharing signal is predicted as label y i The score, Represents the label y i-1 Move to label y i The score, The larger the value, the more label y i-1 The greater the possibility, the conditional probability of predicting the label sequence Y is P(Y / X), and the calculation formula is:

[0124]

[0125] in, represents the true label sequence, Y X Represents all possible label sequences, score(X,Y) represents the correct score of path prediction, Indicates the score of a path among all paths;

[0126] The Viterbi algorithm is used to convert the label sequence Y with the highest score into * As the labeling result of the target BiLSTM+CRF recurrent neural network model, the calculation formula is:

[0127]

[0128] The target transmission unit is used to obtain a target platform transmission address, where the target platform transmission address includes at least one platform transmission address or two or more platform transmission addresses.

[0129] The exchange sharing module is used to obtain the encrypted platform data information and the target platform transmission address, and transmit the encrypted platform data information to the target platform transmission address;

[0130] Specifically, this embodiment includes a data transmission unit, which is used to obtain encrypted platform data information and a target platform transmission address, and transmit the encrypted platform data information to the target platform transmission address; a platform receiving unit, which is used to receive at least one encrypted platform data information. If more than two encrypted platform data information are received, the encrypted platform data information is transmitted platform by platform; a platform sharing unit, which is used to input the encrypted platform data into a shared database for storage. If the corresponding encrypted platform data information needs to be called, the encrypted platform data information is obtained through the shared database; a platform exchange unit, which is used to obtain the exchange requirements of each platform and transmit the encrypted platform data information according to the exchange requirements.

[0131] The transmission feedback module is used to obtain the platform exchange feedback result. If the platform exchange feedback result is that the encrypted platform data information transmission address is incorrect, the platform exchange sharing signal is re-input into the target BiLSTM+CRF recurrent neural network model for recognition.

[0132] Specifically, this embodiment includes a platform feedback unit, which is used to obtain the platform exchange feedback result. If the platform exchange feedback result is that the encrypted platform data information transmission address is incorrect; a data storage unit, which is used to transmit the encrypted platform data information to a shared database for storage if the platform exchange feedback result is that the encrypted platform data information transmission address is correct; and a data identification unit, which is used to re-input the platform exchange shared signal into the target BiLSTM+CRF recurrent neural network model for identification.

[0133] Its beneficial effects are that, through the data acquisition module, it is used to obtain the platform data information and platform exchange sharing signals corresponding to each platform; the data processing module is used to perform data preprocessing on the platform data information to obtain the first platform data information; the data encryption module is used to obtain the first platform data information, encrypt the first platform data information through the AES encryption algorithm, and obtain the encrypted platform data information; the model building module is used to establish the initial BiLSTM+CRF recurrent neural network model based on the RNN recurrent neural network, and use the LeakyReLu activation function to replace the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model to obtain the target The system consists of a standard BiLSTM+CRF recurrent neural network model; an address recognition module, which obtains the platform exchange sharing signal and inputs it into the target BiLSTM+CRF recurrent neural network model for identification, thereby obtaining the target platform transmission address; an exchange sharing module, which obtains the encrypted platform data information and the target platform transmission address and transmits the encrypted platform data information to the target platform transmission address; and a transmission feedback module, which obtains the platform exchange feedback result. If the platform exchange feedback result indicates that the encrypted platform data information transmission address is incorrect, the platform exchange sharing signal is re-input into the target BiLSTM+CRF recurrent neural network model for identification. 1. Significantly Improved Data Transmission Efficiency: By adopting standardized data exchange protocols and interface specifications, and optimizing cross-platform data transmission mechanisms, data transmission time can be significantly shortened, data delays and loss during transmission can be reduced, and the real-time and accuracy of data transmission can be significantly improved. 2. Reduced Data Exchange and Sharing Costs: This system achieves seamless data integration and efficient sharing, reducing conversion and adaptation costs caused by data format mismatches and communication protocol differences. Furthermore, by utilizing advanced technologies such as cloud computing for data storage and processing, local storage and computing resource consumption can be effectively reduced, further reducing operating costs. 3. Improve urban management: As a crucial component of smart cities, the parking lot vehicle management platform's improved data transmission and sharing efficiency will help city managers better understand urban traffic dynamics, optimize traffic resource allocation, and improve urban traffic operation efficiency. At the same time, in-depth mining and analysis of massive parking data can provide strong support for urban planning and traffic policy formulation. 4. Strengthen data security and privacy protection: While improving data transmission efficiency and sharing capabilities, strengthening data security measures to ensure the security and privacy of data during transmission and sharing will help build user trust and protect user rights. This is of great significance for enhancing the reliability and credibility of the entire parking lot vehicle management platform.

[0134] In this embodiment, please refer to Figure 2In the second embodiment of the artificial intelligence management system for cross-platform data exchange and sharing according to the present invention, the data encryption module includes the following units:

[0135] The encryption matrix unit is used to ensure that the 16-byte plaintext, 16-byte ciphertext and wheel key in the AES encryption algorithm are all a 4x4 matrix;

[0136] A round key adding unit is used to perform an XOR operation on each first platform data information in the matrix and the secondary round key, and each subkey is generated by the key;

[0137] a replacement unit, configured to replace each piece of first platform data information with corresponding second platform data information using a lookup table through a nonlinear replacement function;

[0138] A row shift unit, configured to cyclically shift the horizontal columns of each second platform data information in the matrix;

[0139] The column obfuscation unit is used to mix the four second platform data information in each column by using linear transformation to obtain encrypted platform data information.

[0140] Its beneficial effect is to strengthen data security and privacy protection: while improving data transmission efficiency and sharing capabilities, it also strengthens data security measures to ensure the security and privacy of data during transmission and sharing, helping to build user trust and protect user rights. This is of great significance to improving the reliability and credibility of the entire parking lot vehicle management platform.

[0141] In this embodiment, please refer to Figure 3 In the third embodiment of the artificial intelligence management system for cross-platform data exchange and sharing according to the present invention, the model building module includes the following units:

[0142] A model building unit is used to build an initial BiLSTM+CRF recurrent neural network model based on the RNN recurrent neural network, and replace the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model with the LeakyReLu activation function;

[0143] The configuration parameter unit is used to automatically search for the configuration parameters of the initial BiLSTM+CRF recurrent neural network model based on the CondAFP framework and the design space exploration of the genetic algorithm;

[0144] The filter pruning unit is used in the CondAFP framework to perform conditional activation-based model optimization and filter pruning-based model optimization strategies according to the gene expression of each individual chromosome in the initial BiLSTM+CRF recurrent neural network model, thereby obtaining a compressed initial BiLSTM+CRF recurrent neural network model.

[0145] The threshold activation unit is used to activate according to the threshold corresponding to each stage in the gene using entropy perception, predict input samples of different difficulty levels through multiple stages with different computational complexity, and remove unimportant filters to compress the initial BiLSTM+CRF recurrent neural network model;

[0146] Adaptability unit, used in the CondAFP framework to evaluate the adaptability of the corresponding individuals based on the accuracy and FLOPS of the compressed multi-stage model and rank the individuals according to their adaptability;

[0147] The model iteration unit is used to update the chromosome set to obtain the next generation of individuals, and then obtain the target BiLSTM+CRF recurrent neural network model through iteration.

[0148] The beneficial effect is that it enables seamless data integration and efficient sharing, reducing conversion and adaptation costs caused by data format mismatches and communication protocol differences. Furthermore, by utilizing a neural network model for structural improvement and pruning optimization, and applying it to parking lot data, it can effectively identify the target transmission address of the data and improve transmission efficiency.

[0149] See also Figure 4 , a schematic diagram of an embodiment of an artificial intelligence management method for cross-platform data exchange and sharing in an embodiment of the present invention, the method includes the following steps:

[0150] Step 101: Acquire platform data information and platform exchange sharing signals corresponding to each platform;

[0151] Step 102: pre-process the platform data information to obtain first platform data information;

[0152] Step 103: Obtain first platform data information, and encrypt the first platform data information using an AES encryption algorithm to obtain encrypted platform data information;

[0153] Step 104: Establish an initial BiLSTM+CRF recurrent neural network model based on the RNN recurrent neural network, and use the LeakyReLu activation function to replace the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model to obtain a target BiLSTM+CRF recurrent neural network model;

[0154] Step 105: Obtain the platform exchange sharing signal, input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for recognition, and obtain the target platform transmission address;

[0155] Step 106: Obtain the encrypted platform data information and the target platform transmission address, and transmit the encrypted platform data information to the target platform transmission address;

[0156] Step 107: Obtain the platform exchange feedback result. If the platform exchange feedback result is that the encrypted platform data information transmission address is incorrect, re-input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for recognition.

[0157] The beneficial effect lies in the fact that the parking lot vehicle management platform, as a key component of smart cities, improves the efficiency of data transmission and sharing, helping city managers better understand urban traffic dynamics, optimize transportation resource allocation, and improve urban transportation efficiency. Furthermore, through in-depth mining and analysis of massive parking data, it can provide strong support for urban planning and transportation policy formulation.

[0158] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention, and such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence management system applied to cross-platform data exchange and sharing, characterized by: The artificial intelligence management system for data exchange and sharing includes a data acquisition module, a data processing module, a data encryption module, a model building module, an address recognition module, an exchange and sharing module, and a transmission feedback module: A data acquisition module is used to obtain platform data information corresponding to each platform and platform exchange sharing signals; A data processing module, used for performing data preprocessing on the platform data information to obtain first platform data information; A data encryption module is used to obtain the first platform data information, and encrypt the first platform data information by using an AES encryption algorithm to obtain encrypted platform data information; A model building module is used to build an initial BiLSTM+CRF recurrent neural network model based on the RNN recurrent neural network, and replace the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model with a LeakyReLu activation function to obtain a target BiLSTM+CRF recurrent neural network model; An address recognition module is used to obtain the platform exchange sharing signal, input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for recognition, and obtain the target platform transmission address; An exchange sharing module, used to obtain the encrypted platform data information and the target platform transmission address, and transmit the encrypted platform data information to the target platform transmission address; A transmission feedback module is used to obtain a platform exchange feedback result. If the platform exchange feedback result is an encrypted platform data information transmission address error, the platform exchange sharing signal is re-input into the target BiLSTM+CRF recurrent neural network model for identification; The data acquisition module includes a platform data acquisition unit and an exchange signal acquisition unit: A platform data acquisition unit, used to acquire platform data information corresponding to each platform, wherein the platform data information is operation data in the parking lot vehicle management platform; The exchange signal acquisition unit is used to acquire data acquisition instructions from each platform system and generate a platform exchange sharing signal based on the data acquisition instructions; the data acquisition instructions include voice input instructions, text input instructions, and vehicle image input instructions.

2. The artificial intelligence management system for cross-platform data exchange and sharing according to claim 1, characterized in that: The data processing module includes a data screening unit, a vectorization unit, a normalization unit, and a data classification unit: A data screening unit is used to obtain platform data information corresponding to each platform, and screen the platform features of the platform data information by using feature engineering to obtain a platform feature label library; A vectorization unit, used for vectorizing the text information of the platform feature tag library by one-hot encoding, converting it into a digital vector using an encoding method, and obtaining a platform text feature tag library; A normalization unit, used for normalizing the platform text feature label library to obtain a target platform text feature label library, wherein the target platform text feature label library conforms to a standard normal distribution; The data classification unit is used to classify the target platform text feature tag library through the FCM fuzzy clustering algorithm to obtain the first platform data information.

3. The artificial intelligence management system for cross-platform data exchange and sharing according to claim 1, characterized in that: The data encryption module includes an encryption matrix unit, a round key adding unit, a substitution unit, a row shift unit, and a column confusion unit: The encryption matrix unit is used to determine that the 16-byte plaintext, 16-byte ciphertext and wheel key in the AES encryption algorithm are all a 4x4 matrix; A round key adding unit, used for performing an XOR operation on each first platform data information in the matrix and the secondary round key, and each subkey is generated by the key; A replacement unit, used to replace each first platform data information with corresponding second platform data information by using a lookup table through a non-linear replacement function; A row shift unit, used for cyclically shifting the horizontal columns of each second platform data information in the matrix; The column confusion unit is used to mix the four second platform data information of each column by linear conversion to obtain encrypted platform data information.

4. The artificial intelligence management system for cross-platform data exchange and sharing according to claim 1, characterized in that: The model building module includes a model building unit, a configuration parameter unit, a filter pruning unit, a threshold activation unit, an adaptability unit, and a model iteration unit: A model building unit, used to build an initial BiLSTM+CRF recurrent neural network model based on the RNN recurrent neural network, and replace the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model with a LeakyReLu activation function; Configuration parameter unit, used to automatically search for the configuration parameters of the initial BiLSTM+CRF recurrent neural network model based on the CondAFP framework and the design space exploration using genetic algorithms; The filter pruning unit is used in the CondAFP framework to execute the model optimization strategy based on conditional activation and the model optimization strategy based on filter pruning according to the gene expression of each individual chromosome in the initial BiLSTM+CRF recurrent neural network model, and obtain the compressed initial BiLSTM+CRF recurrent neural network model; Threshold activation unit, which is used to activate according to the threshold corresponding to each stage in the gene using entropy perception, predict input samples of different difficulty through multiple stages with different computational complexity, and remove unimportant filters to compress the initial BiLSTM+CRF recurrent neural network model; The adaptability unit is used in the CondAFP framework to evaluate the adaptability of the corresponding individuals according to the accuracy and FLOPS of the compressed multi-stage model, and to rank the individuals according to their adaptability; The model iteration unit is used to update the chromosome set to obtain the next generation of individuals, and obtain the target BiLSTM+CRF recurrent neural network model after iteration.

5. The artificial intelligence management system for cross-platform data exchange and sharing according to claim 1, characterized in that: The address recognition module includes a signal acquisition unit, a signal recognition unit, and a target transmission unit: A signal acquisition unit, used to acquire the platform exchange sharing signal, wherein the platform exchange sharing signal at least includes a platform information acquisition signal and a platform information output signal; A signal recognition unit, used for inputting the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for recognition; The BiLSTM layer in the target BiLSTM+CRF recurrent neural network model includes at least three gating units, namely, a forget gate, an input gate, and an output gate. The calculation formula is: in, , , , Indicated in The forget gate, input gate, output gate and memory cell at each moment; , , , Represents the weight matrix corresponding to different control gates; , , , represents the bias vector; represents the intermediate state of the input, express The input vector at time instant, express The output result at the moment, represents the LeakyReLu activation function, represents the hyperbolic tangent activation function, Represents the dot product operator; The CRF layer in the target BiLSTM+CRF recurrent neural network model performs joint probability analysis on the label sequence by analyzing the relationship between adjacent labels. The analysis process is as follows: Assume that P is the output matrix of the BiLSTM layer, n is the number of exchanged shared signals, if represents the input exchange shared signal sequence, represents the corresponding output label sequence, then the input sequence Corresponding prediction sequence Score The calculation formula is: in, Indicates The exchange shared signal is predicted as the label The score, Representation Tags Transfer to label The score, The larger the value, the more labels The greater the probability, the more likely the predicted label sequence The conditional probability of , the calculation formula is: in, represents the true label sequence, represents all possible label sequences, represents the correct score of path prediction, Indicates the score of a path among all paths; The Viterbi algorithm is used to convert the label sequence with the highest score into As the labeling result of the target BiLSTM+CRF recurrent neural network model, the calculation formula is: The target transmission unit is used to obtain a target platform transmission address, where the target platform transmission address includes at least one platform transmission address and more than two platform transmission addresses.

6. The artificial intelligence management system for cross-platform data exchange and sharing according to claim 1, characterized in that: The exchange and sharing module includes a data transmission unit, a platform receiving unit, a platform sharing unit, and a platform exchange unit: A data transmission unit, used for acquiring the encrypted platform data information and the target platform transmission address, and transmitting the encrypted platform data information to the target platform transmission address; The platform receiving unit is used to receive at least one encrypted platform data information. If more than two encrypted platform data information are received, the encrypted platform data information is transmitted one by one on the platform. The platform sharing unit is used to input the encrypted platform data into the shared database for storage. If the corresponding encrypted platform data information needs to be called, the encrypted platform data information is obtained through the shared database; The platform exchange unit is used to obtain the exchange requirements of each platform and transmit the encrypted platform data information according to the exchange requirements.

7. The artificial intelligence management system for cross-platform data exchange and sharing according to claim 1, characterized in that: The transmission feedback module includes a platform feedback unit, a data storage unit, and a data identification unit: A platform feedback unit, used to obtain a platform exchange feedback result, if the platform exchange feedback result is an encrypted platform data information transmission address error; A data storage unit, configured to transmit the encrypted platform data information to a shared database for storage if the platform exchange feedback result indicates that the encrypted platform data information transmission address is correct; A data identification unit is used to re-input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for identification.

8. An artificial intelligence management method for cross-platform data exchange and sharing, characterized in that: The artificial intelligence management method comprises the following steps: Obtain platform data information corresponding to each platform and platform exchange sharing signals; The platform data information is the operational data in the parking lot vehicle management platform; Performing data preprocessing on the platform data information to obtain first platform data information; Acquire the first platform data information, and encrypt the first platform data information using an AES encryption algorithm to obtain encrypted platform data information; An initial BiLSTM+CRF recurrent neural network model is established based on the RNN recurrent neural network, and the ReLu activation function of the initial BiLSTM+CRF recurrent neural network model is replaced with the LeakyReLu activation function to obtain a target BiLSTM+CRF recurrent neural network model; Obtain the platform exchange sharing signal, input the platform exchange sharing signal into the target BiLSTM+CRF recurrent neural network model for identification, and obtain the target platform transmission address; Acquire the encrypted platform data information and the target platform transmission address, and transmit the encrypted platform data information to the target platform transmission address; The platform exchange feedback result is obtained. If the platform exchange feedback result is that the encrypted platform data information transmission address is wrong, the platform exchange sharing signal is re-input into the target BiLSTM+CRF recurrent neural network model for identification.

9. The artificial intelligence management method for cross-platform data exchange and sharing according to claim 8, characterized in that: The artificial intelligence management method comprises the following steps: Obtain the platform data information corresponding to each platform, use feature engineering to filter the platform features of the platform data information, and obtain the platform feature label library; Vectorizing the text information of the platform feature tag library by one-hot encoding, converting it into a digital vector using an encoding method, and obtaining a platform text feature tag library; Normalizing the platform text feature label library to obtain a target platform text feature label library, wherein the target platform text feature label library conforms to a standard normal distribution; The target platform text feature tag library is classified by using the FCM fuzzy clustering algorithm to obtain the first platform data information.

Images