Big data asset management system

Abstract

The invention discloses a big data asset management system, and the system comprises an asset information obtaining module which obtains asset information through an electronic tag technology; the actual consumption acquisition module is used for counting the actual consumption of the whole staff in the target time period; the recovery state judgment module is used for judging whether the recovery state of the assets is abnormal or not by comparing the actual consumption of the whole staff with the actual recovery amount of the whole staff; the predicted consumption acquisition module is used for combining RNN and LSTM to construct a prediction model for predicting the actual consumption of the whole personnel, collecting internal historical data and external data sources of the asset management system to carry out prediction model training, and predicting the predicted consumption of the whole personnel; the consumption state judgment module is used for judging whether the consumption state of the assets is abnormal or not by comparing the predicted consumption of the whole staff with the actual consumption of the whole staff; and the credit evaluation module is used for constructing a credit evaluation model based on multi-dimensional data, comprehensively evaluating the credit condition of the user, predicting the credit of the user and identifying a potential high-risk user.

Description

Technical Field

[0001] The present invention relates to the technical field of asset management, and particularly to a big data asset management system. Background Art

[0002] With the progress of technology and the popularization of the Internet, human society has gradually entered the digital economy era. In this era, data is regarded as a new and important production factor, on a par with land, labor, capital, and technology, and has a profound impact on economic and social development. The monetization of data assets means that data resources have changed from natural resources to economic assets, which will directly affect the input and output of companies. In recent years, the state has continuously promoted the construction of the data factor market and introduced a series of relevant policies and regulations, providing strong institutional guarantees for big data asset management. For example: The Twenty Articles on Data: The "Opinions on Building a Data Basic System to Better Play the Role of Data Factors" (referred to as the "Twenty Articles on Data") issued in 2022 aims to conduct a comprehensive institutional design for data factors at the national level, promote the transformation of data resources into data assets, and provide a solid institutional guarantee for the high-quality development of the digital economy.

[0003] After retrieval, the invention patent with the Chinese patent number CN117522168A discloses an asset management method and system based on big data, belonging to the technical field of asset management. The method includes the following steps: Based on the electronic tags on each target asset, obtain asset information, statistically obtain the total actual consumption and total actual recovery within the target period according to the asset information, and compare them to judge the asset recovery status; when the asset recovery status is abnormal, screen out high-credit users, and estimate the total predicted consumption based on the personal actual consumption of the high-credit users; compare the total predicted consumption and the total actual consumption to judge the asset consumption status; The present invention can timely determine whether enterprise assets are excessively and illegally occupied while ensuring the needs of employees, protecting the rights and interests of the enterprise.

[0004] However, in the process of using the above patent, the accuracy of the total predicted consumption depends on the accuracy and reliability of the prediction model. However, due to the influence of various factors, such as market changes, user demand changes, etc., the prediction results are inaccurate. Therefore, a big data asset management system is proposed. Summary of the Invention

[0005] The purpose of the present invention is to solve the drawback in the prior art that the accuracy of the total predicted consumption depends on the accuracy and reliability of the prediction model. However, due to the influence of various factors, such as market changes, user demand changes, etc., the prediction results are inaccurate, and a big data asset management system is proposed.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A big data asset management system, comprising:

[0008] An asset information acquisition module: obtaining asset information through electronic tag technology, where the asset information is the basis for the actual consumption acquisition module to perform consumption statistics;

[0009] An actual consumption acquisition module: used to count the actual consumption of all employees within a target period;

[0010] A recovery status judgment module: judging whether the recovery status of the asset is abnormal by comparing the actual consumption of all employees and the actual recovery amount of all employees, and taking management measures;

[0011] A predicted consumption acquisition module: combining RNN and LSTM to construct a prediction model for predicting the actual consumption of all employees, collecting historical data inside the asset management system and external data sources for prediction model training, estimating the predicted consumption of all employees, and triggering the function of the predicted consumption acquisition module when the recovery status judgment module determines that the asset recovery status is abnormal. The predicted consumption acquisition module estimates the predicted consumption of all employees based on the personal actual consumption of high-credit users and in combination with the prediction model;

[0012] A consumption status determination module: judging whether the consumption status of the asset is abnormal by comparing the predicted consumption of all employees and the actual consumption of all employees, and generating corresponding prompt information and management measures. When the consumption status determination module determines that the asset consumption status is abnormal, it refers to the result of the credit evaluation module and pays special attention to and manages low-credit users;

[0013] A credit evaluation module: constructing a credit evaluation model based on multi-dimensional data, comprehensively evaluating the credit status of users and predicting the credit of users, and identifying potential high-risk users.

[0014] The above technical solutions further include:

[0015] Furthermore, the asset information acquisition module includes an electronic tag reading unit, a data storage unit, and a data preprocessing unit. The electronic tag reading unit is responsible for reading the electronic tag information on the asset to ensure the accuracy and real-time nature of the data. The data storage unit is used to store the asset information obtained from the electronic tag reading unit. The data preprocessing unit preprocesses the stored asset information, such as data cleaning and format conversion, to ensure the quality and consistency of the data. After the electronic tag reading unit reads the information from the electronic tag on the asset, it transfers the electronic tag information to the data storage unit for storage. The asset information stored in the data storage unit will be read and preprocessed by the data preprocessing unit, and the preprocessed asset information is transferred to the actual consumption acquisition module as the basic data for consumption quantity statistics.

[0016] Furthermore, the predicted consumption acquisition module includes a data collection and integration unit, a model design and training unit, a high-credit user determination unit, a predicted consumption calculation unit, and a data output and interface unit. The data collection and integration unit is responsible for collecting historical data within the asset management system, such as the allocation situation, consumption quantity, and recovery quantity of assets, as well as external data sources, such as market trends, seasonal changes, and policy changes. The historical data within the asset management system and the external data sources will be integrated and used for the training of the prediction model. The model design and training unit designs a hybrid prediction model containing RNN and LSTM layers based on RNN and LSTM technologies. The model design and training unit is responsible for building the model, adjusting the parameters, and training the model using the collected data. When the recovery status judgment module determines that the asset recovery status is abnormal, the high-credit user determination unit will, based on the results of the user credit evaluation module, screen out high-credit users whose credit exceeds a preset threshold. The personal actual consumption of the high-credit users will be used for subsequent predicted consumption calculation. The predicted consumption calculation unit calculates the predicted consumption for all employees based on the trained prediction model and the personal actual consumption of high-credit users. This is the core function of the predicted consumption acquisition module and provides important prediction data for asset management. The data output and interface unit outputs the calculated predicted consumption for all employees to the consumption status judgment module for subsequent consumption status judgment and management measure implementation.

[0017] Furthermore, the model design and training unit conducts model design, including the following steps:

[0018] Model architecture:

[0019] The prediction model consists of an RNN layer and an LSTM layer. The RNN layer is used to capture short-term dependencies in the data, and the LSTM layer is used to capture long-term dependencies.

[0020] RNN layer: Receives the input sequence X = (x 1 , x 2 ,..., x T ), where x t represents the input at the t-th time step. The RNN layer passes information through the hidden state h t . The calculation formula for h t is

[0021] h t = σ(W hh h t-1 + W xh x t + b h )

[0022] where σ is the activation function, W hh and W xh are weight matrices, and b h is the bias term;

[0023] LSTM layer: Receives the output of the RNN layer as input and controls the flow of information through the forget gate, input gate, and output gate. The hidden state c t and output h' t of the LSTM layer are calculated as follows: Forget gate: f t = σ(W f · [h' t-1 , x t + b f )

[0024] Input gate: i t = σ(W i · [h' t-1 , x t + b i )

[0025] Candidate cell state:

[0026] Cell state update:

[0027] Output gate: o t = σ(W o · [h' t-1 , x t + b o )

[0028] Hidden state output: h' t = o t · tanh(c t )

[0029] where W hh , W rh, W f , W i , W c and W o are weight matrices, and b h , b f , b i , b c and b o are bias terms;

[0030] Parameter initialization:

[0031] The weight matrices W hh , W xh , W f , W i , W c and W o in the model, as well as the bias term b h , b f , b i , b c and b o are randomly initialized using a normal distribution or a uniform distribution;

[0032] Loss function:

[0033] To train the model, a loss function is defined, and the calculation formula of the loss function is

[0034]

[0035] where n is the number of samples, y i is the actual value of the i-th sample, is the predicted value of the i-th sample.

[0036] Furthermore, the model design and training unit performs model training, including the following steps:

[0037] Data preparation: Collect and integrate the historical data inside the asset management system and external data sources to form a training dataset, where the data includes the input sequence X and the corresponding output sequence Y;

[0038] Forward propagation: Input the training data into the model and perform forward propagation through the RNN layer and the LSTM layer to obtain the predicted value

[0039] Calculate the loss: Use the loss function to calculate the difference between the predicted value and the actual value Y to obtain the loss value;

[0040] Backward propagation and parameter update: Calculate the gradient of the loss value with respect to the model parameters through the backward propagation algorithm and use Adam to update the model parameters;

[0041] Iterative Training:

[0042] Repeat the processes of forward propagation, loss calculation, and backpropagation with parameter update until the preset number of iterations is reached or the loss value converges.

[0043] Furthermore, the credit assessment module includes a data integration unit, a credit score calculation unit, a high-risk user identification unit, and a credit management unit. The data integration unit is responsible for obtaining data from various data sources, cleaning, transforming, and integrating it, and passing the integrated data to the credit score calculation unit and the high-risk user identification unit. The credit score calculation unit calculates the credit score of the user. The credit score calculation unit receives the integrated data passed by the data integration unit, calculates to generate a credit score, and passes the score result to the high-risk user identification unit and the credit management unit. The high-risk user identification unit receives the credit score passed by the credit score calculation unit, conducts risk analysis to generate a list of high-risk users, and passes the result to the credit management unit. The credit management unit takes corresponding management measures based on the user's credit score and risk level. The credit management unit receives the list of high-risk users passed by the high-risk user identification unit and the credit score passed by the credit score calculation unit, makes a decision to generate corresponding management instructions, such as restricting permissions, increasing thresholds, etc.

[0044] Furthermore, the specific steps for the credit score calculation unit to calculate the credit score are as follows:

[0045] Data preprocessing: Clean and organize the original data, including missing value handling, outlier detection and handling, data type conversion, etc.;

[0046] Feature selection: Select the features that have an impact on the credit score, and select them based on statistical methods, machine learning algorithms, or expert experience;

[0047] The features that may be selected include age, income, occupation, historical borrowing records, repayment status, etc.;

[0048] Feature scaling: Use standardization for scaling to make different features have the same weight in the model. The standardization formula is

[0049]

[0050] where x is the original data, μ is the mean, and σ is the standard deviation;

[0051] Model construction: Build a credit score model based on the selected features and logistic regression;

[0052] Model Evaluation: Use the test set data for verification, calculate metrics such as accuracy, recall, F1-score, etc., to evaluate the accuracy and reliability of the model. If the model has a high accuracy on the test set and the false positive rate and false negative rate are both within an acceptable range, the model is considered effective;

[0053] Credit Score Calculation: Input the user's feature data into the credit scoring model to obtain the prediction result, and then convert it into a credit score according to the score mapping table;

[0054] Result Interpretation and Reporting: Explain the meaning of the credit score, the calculation method, and the key factors affecting the score. The report may include the user's credit score, score level, positive and negative factors affecting the score, etc.

[0055] Furthermore, for the model construction, the specific steps are as follows:

[0056] Data Preparation: Collect and organize the dataset for model training, including the selected features and corresponding credit scores. Extract data from the database and perform preprocessing tasks such as missing value handling, outlier detection and handling, and data cleaning;

[0057] Feature Engineering: Use standardization to scale the features so that they have the same weight in the model. At the same time, select or eliminate some features as needed;

[0058] Model Training: Use the processed data of logistic regression for training to obtain the credit scoring model. The logistic regression prediction formula is

[0059]

[0060] where Y is the target variable, X is the feature vector, β 0 ,β 1 ,...,β n are the parameters of the model, which need to be estimated through the training process. Initialize the model parameters, select the logarithmic loss function and the gradient descent method, and iteratively update the model parameters until the convergence condition (such as the loss function value no longer decreases significantly) is reached, and find a set of optimal parameters β to make the prediction accuracy of the model on the training set the highest;

[0061] Parameter Estimation: After training, obtain the estimated values of the model parameters β, and the parameters reflect the influence degree of each feature on the target variable Y;

[0062] Model Evaluation: Use the test set data for verification, calculate metrics such as accuracy, recall, F1-score, etc. At the same time, tools such as confusion matrix and ROC curve can be used for visualization analysis;

[0063] Model Deployment: Integrate the model into the credit scoring calculation unit.

[0064] The present invention has the following beneficial effects:

[0065] In the present invention, by integrating data resources from different channels, including data within the asset management system and external data sources, rich and comprehensive data support is provided for the prediction model. By combining RNN and LSTM, a prediction model for predicting the actual consumption of all employees is constructed, learning complex non-linear relationships from the data, and adapting to consumption patterns under different time periods and conditions, which can more effectively capture long-term dependencies in time series data and improve the accuracy and stability of predictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a system block diagram of a big data asset management system proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0067] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0068] Please refer to Figure 1 As shown, the present invention is a big data asset management system, including:

[0069] Asset information acquisition module: Through the electronic tag technology, asset information is acquired, which is the basis for the actual consumption acquisition module to perform consumption statistics.

[0070] Actual consumption acquisition module: Used to count the actual consumption of all employees within the target time period.

[0071] Recovery status judgment module: By comparing the actual consumption of all employees and the actual recovery amount of all employees, it judges whether the recovery status of the asset is abnormal and takes management measures.

[0072] Predicted consumption acquisition module: Combining RNN and LSTM, a prediction model for predicting the actual consumption of all employees is constructed. Historical data within the asset management system and external data sources are collected for prediction model training to estimate the predicted consumption of all employees. When the recovery status judgment module determines that the asset recovery status is abnormal, the function of the predicted consumption acquisition module is triggered. The predicted consumption acquisition module estimates the predicted consumption of all employees based on the personal actual consumption of high-credit users and in combination with the prediction model.

[0073] Consumption status determination module: By comparing the predicted consumption of all employees with the actual consumption of all employees, it determines whether the consumption status of the asset is abnormal, and generates corresponding prompt messages and management measures. When the consumption status determination module determines that the asset consumption status is abnormal, it refers to the results of the credit assessment module and pays special attention to and manages low-credit users;

[0074] Credit assessment module: Build a credit assessment model based on multi-dimensional data, comprehensively evaluate the user's credit status, predict the user's credit, and identify potential high-risk users.

[0075] In one embodiment, for the above asset information acquisition module, the asset information acquisition module includes an electronic tag reading unit, a data storage unit, and a data preprocessing unit. The electronic tag reading unit is responsible for reading the electronic tag information on the asset to ensure the accuracy and timeliness of the data. The data storage unit is used to store the asset information obtained from the electronic tag reading unit. The data preprocessing unit preprocesses the stored asset information, such as data cleaning, format conversion, etc., to ensure the quality and consistency of the data. After the electronic tag reading unit reads the information from the electronic tag on the asset, it transmits the electronic tag information to the data storage unit for storage. The asset information stored in the data storage unit will be read and preprocessed by the data preprocessing unit. The preprocessed asset information is transmitted to the actual consumption acquisition module as the basic data for consumption volume statistics.

[0076] In one embodiment, for the above-mentioned predicted consumption acquisition module, the predicted consumption acquisition module includes a data collection and integration unit, a model design and training unit, a high-credit user determination unit, a predicted consumption calculation unit, and a data output and interface unit. The data collection and integration unit is responsible for collecting historical data within the asset management system, such as asset allocation, consumption, recycling, etc., as well as external data sources, such as market trends, seasonal changes, policy changes and other information. The historical data within the asset management system and the external data sources will be integrated and used for training the prediction model. The model design and training unit designs a hybrid prediction model containing RNN and LSTM layers based on RNN and LSTM technology. The model design and training unit is responsible for the model. The module builds, adjusts parameters and uses the collected data to train the model. When the recovery status judgment module determines that the asset recovery status is abnormal, the high-credit user judgment unit will screen out high-credit users whose credit exceeds the preset threshold based on the results of the user credit evaluation module. The personal actual consumption of high-credit users will be used for subsequent predicted consumption calculations. The predicted consumption calculation unit calculates the predicted consumption of all employees based on the trained prediction model and the personal actual consumption of high-credit users. This is the core function of the predicted consumption acquisition module, which provides important prediction data for asset management. The data output and interface unit outputs the calculated predicted consumption of all employees to the consumption status judgment module for subsequent consumption status judgment and management measures.

[0077] In one embodiment, for the above-mentioned model design and training unit, the model design and training unit performs model design, including the following steps:

[0078] Model Architecture:

[0079] The prediction model consists of an RNN layer and an LSTM layer. The RNN layer is used to capture short-term dependencies in the data, and the LSTM layer is used to capture long-term dependencies.

[0080] RNN layer: receives input sequence X = (x 1 , x 2 , ..., x T ), where x t Represents the input of the tth time step, and the RNN layer passes the hidden state h t Transmit information, t The calculation formula is

[0081] h t =σ(W hh h t-1 +W xh x t +b h )

[0082] Among them, σ is the activation function, Whh and W xh is the weight matrix, and b h is the bias term;

[0083] LSTM layer: Receives the output of the RNN layer as input, and controls the flow of information through the forget gate, input gate, and output gate. The hidden state c t and the output h' t are calculated as follows

[0084] Forget gate: f t = σ(W f · [h', t-1 x t + b f )

[0085] Input gate: i t = σ(W i · [h', t-1 x t + b i )

[0086] Candidate cell state:

[0087] Cell state update:

[0088] Output gate: o t = σ(W o · [h', t-1 xt] + b o )

[0089] Hidden state output: h' t = o t · tanh(c t )

[0090] where W hh , W zh , W f , W i , W c and W o are the weight matrices, and b h , b f , b i , b c and b o are the bias terms;

[0091] Parameter initialization:

[0092] The weight matrices W hh , W xh , W f , W i , Wc and W o and the bias term b h 、b f 、b i 、b c and b o Use normal distribution or uniform distribution for random initialization;

[0093] Loss function:

[0094] To train the model, define the loss function. The calculation formula of the loss function is

[0095]

[0096] where n is the number of samples, y i is the actual value of the i-th sample, is the predicted value of the i-th sample.

[0097] In one embodiment, for the above model design and training unit, the model design and training unit performs model training, including the following steps:

[0098] Data preparation: Collect and integrate the historical data inside the asset management system and external data sources to form a training dataset. The data includes the input sequence X and the corresponding output sequence Y;

[0099] Forward propagation: Input the training data into the model and perform forward propagation through the RNN layer and the LSTM layer to obtain the predicted value

[0100] Calculate the loss: Use the loss function to calculate the difference between the predicted value and the actual value Y to obtain the loss value;

[0101] Backward propagation and parameter update: Calculate the gradient of the loss value with respect to the model parameters through the backward propagation algorithm and use Adam to update the model parameters;

[0102] Iterative training:

[0103] Repeat the processes of forward propagation, calculating the loss, and backward propagation and parameter update until the preset number of iterations is reached or the loss value converges;

[0104] Suppose there is an asset management system that contains historical consumption and recovery data of assets. Take these data as the input sequence X and the consumption at the next time step as the output sequence Y. Then, train the hybrid prediction model according to the above steps. Before training the model, it is necessary to preprocess the data, including data cleaning, normalization, etc. This helps to improve the training efficiency and prediction accuracy of the model; during the training process, it is observed that the loss value gradually decreases, indicating that the prediction ability of the model is gradually improving. After training is completed, it is necessary to evaluate the model to verify its performance. Commonly used evaluation metrics include accuracy, recall rate, F1 score, etc. In addition, charts such as learning curves and loss curves can also be drawn to observe the training process of the model. When training is completed, use the trained model to predict new data to obtain the predicted consumption of all employees.

[0105] In one embodiment, for the above-mentioned credit assessment module, the credit assessment module includes a data integration unit, a credit score calculation unit, a high-risk user identification unit, and a credit management unit. The data integration unit is responsible for obtaining data from various data sources, cleaning, transforming, and integrating it, and passing the integrated data to the credit score calculation unit and the high-risk user identification unit. The credit score calculation unit calculates the credit score of the user. The credit score calculation unit receives the integrated data passed by the data integration unit, calculates to generate a credit score, and passes the score result to the high-risk user identification unit and the credit management unit. The high-risk user identification unit receives the credit score passed by the credit score calculation unit, conducts risk analysis to generate a list of high-risk users, and passes the result to the credit management unit. The credit management unit takes corresponding management measures according to the user's credit score and risk level. The credit management unit receives the list of high-risk users passed by the high-risk user identification unit and the credit score passed by the credit score calculation unit, makes a decision to generate corresponding management instructions, such as restricting permissions and increasing thresholds.

[0106] In one embodiment, for the above-mentioned credit score calculation unit, the specific steps of credit score calculation by the credit score calculation unit are as follows:

[0107] Data preprocessing: Clean and organize the original data, including missing value handling, outlier detection and handling, data type conversion, etc.;

[0108] Feature selection: Select features that have an impact on the credit score, and select them based on statistical methods, machine learning algorithms, or expert experience;

[0109] Possible features to be selected include age, income, occupation, historical borrowing records, repayment status, etc.;

[0110] Feature scaling: Use standardization for scaling to make different features have the same weight in the model. The standardization formula is

[0111]

[0112] Among them, x is the original data, μ is the mean, and σ is the standard deviation;

[0113] Model construction: Build a credit scoring model based on the selected features and logistic regression;

[0114] Model evaluation: Use the test set data for verification, calculate metrics such as accuracy, recall, and F1-score to evaluate the accuracy and reliability of the model. If the accuracy of the model on the test set is high and both the false positive rate and false negative rate are within an acceptable range, the model is considered effective;

[0115] Credit scoring calculation: Input the feature data of the user into the credit scoring model to obtain the prediction result, and then convert it into a credit score according to the score mapping table;

[0116] Result interpretation and reporting: Explain the meaning of the credit score, calculation method, and key factors affecting the score. The report may include the user's credit score, score level, positive and negative factors affecting the score, etc.;

[0117] First, preprocess the data and select features, retaining three features: age, income, and the number of historical borrowings.

[0118] Then, standardize the features to ensure they have the same weight in the model.

[0119] Next, we use the training dataset to train the logistic regression model and obtain the estimated value of the model parameter β.

[0120] Finally, we input the test dataset into the model for prediction and calculate metrics such as the accuracy, recall, and F1-score of the model. The results show that the model performs well on the test set and has high prediction accuracy.

[0121] Based on the prediction results of the model and the score mapping table, we can calculate the credit score for each user and use it in subsequent processes such as loan approval and risk management.

[0122] In one embodiment, for the above model construction, the specific steps of model construction are as follows:

[0123] Data preparation: Collect and organize the dataset for model training, including the selected features and corresponding credit scores, extract data from the database, perform preprocessing work such as missing value handling, outlier detection and handling, and data cleaning;

[0124] Feature Engineering: Use standardization to scale features so that they have the same weight in the model. At the same time, select or eliminate some features as needed;

[0125] Model Training: Use the processed data of logistic regression for training to obtain a credit scoring model. The logistic regression prediction formula is

[0126]

[0127] where Y is the target variable, X is the feature vector, and β 0 , β 1 ,.., β n are the parameters of the model, which need to be estimated through the training process. Initialize the model parameters, select the logarithmic loss function and the gradient descent method, and iteratively update the model parameters until the convergence condition is reached (such as the loss function value no longer decreases significantly), and find a set of optimal parameters β to make the model have the highest prediction accuracy on the training set;

[0128] Parameter Estimation: After training, obtain the estimated values of the model parameters β. The parameters reflect the influence degree of each feature on the target variable Y;

[0129] Model Evaluation: Use the test set data for verification, and calculate metrics such as accuracy, recall rate, and F1 score. At the same time, tools such as confusion matrix and ROC curve can be used for visual analysis;

[0130] Model Deployment: Integrate the model into the credit scoring calculation unit.

[0131] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A big data asset management system, characterized in that: include: Asset information acquisition module: acquires asset information through electronic tag technology, and the asset information is the basis for the actual consumption acquisition module to perform consumption statistics; Actual consumption acquisition module: used to count the actual consumption of all employees within the target period; Recycling status judgment module: By comparing the actual consumption of all employees with the actual recycling of all employees, it is determined whether the recycling status of assets is abnormal and management measures are taken; Predicted consumption acquisition module: Combine RNN and LSTM to build a prediction model for predicting the actual consumption of all employees, collect historical data within the asset management system and external data sources to train the prediction model, and estimate the predicted consumption of all employees. When the recovery status judgment module determines that the asset recovery status is abnormal, the function of the predicted consumption acquisition module is triggered. The predicted consumption acquisition module estimates the predicted consumption of all employees based on the personal actual consumption of high-credit users and the prediction model; Consumption status determination module: by comparing the predicted consumption of all employees with the actual consumption of all employees, it determines whether the consumption status of the asset is abnormal, and generates corresponding prompt information and management measures. When the consumption status determination module determines that the asset consumption status is abnormal, it refers to the results of the credit evaluation module and pays special attention and management to low-credit users; Credit assessment module: Build a credit assessment model based on multi-dimensional data, comprehensively evaluate the user's credit status and predict the user's credit, and identify potential high-risk users.

2. A big data asset management system according to claim 1, characterized in that: The asset information acquisition module includes an electronic tag reading unit, a data storage unit and a data preprocessing unit. The electronic tag reading unit is responsible for reading the electronic tag information on the asset. The data storage unit is used to store the asset information obtained from the electronic tag reading unit. The data preprocessing unit preprocesses the stored asset information. After reading the information from the electronic tag on the asset, the electronic tag reading unit passes the electronic tag information to the data storage unit for storage. The asset information stored in the data storage unit will be read and preprocessed by the data preprocessing unit. The preprocessed asset information is passed to the actual consumption acquisition module as basic data for consumption statistics.

3. A big data asset management system according to claim 1, characterized in that: The predicted consumption acquisition module includes a data collection and integration unit, a model design and training unit, a high-credit user determination unit, a predicted consumption calculation unit, and a data output and interface unit. The data collection and integration unit is responsible for collecting historical data within the asset management system and external data sources. The historical data within the asset management system and the external data sources will be integrated and used for training the prediction model. The model design and training unit designs a hybrid prediction model including RNN and LSTM layers based on RNN and LSTM technologies. The model design and training unit is responsible for model construction, parameter adjustment, and model training using the collected data. When the recovery status judgment module determines that the asset recovery status is abnormal, the high-credit user determination unit will screen out high-credit users whose credit exceeds a preset threshold based on the results of the user credit evaluation module. The personal actual consumption of the high-credit user will be used for subsequent predicted consumption calculations. The predicted consumption calculation unit calculates the predicted consumption of all employees based on the trained prediction model and the personal actual consumption of the high-credit user. The data output and interface unit outputs the calculated predicted consumption of all employees to the consumption status judgment module.

4. A big data asset management system according to claim 3, characterized in that: The model design and training unit performs model design, including the following steps: Model Architecture: The prediction model consists of an RNN layer and an LSTM layer, wherein the RNN layer is used to capture short-term dependencies in the data, and the LSTM layer is used to capture long-term dependencies; RNN layer: receives input sequence X = (x1, x2, ..., x T ), where x t Represents the input of the t-th time step, and the RNN layer passes through the hidden state h t Transmit information, t The calculation formula is h t =σ(W hh h t-1 +W xh x t +b h ) Among them, σ is the activation function, W hh and W xh is the weight matrix, b h is the bias term; LSTM layer: receives the output of the RNN layer as input, controls the flow of information through the forget gate, input gate, and output gate. The hidden state c of the LSTM layer t and output h′ t The calculation formula is Forget gate: f t =σ(W f ·[h′ t-1 , x t ]+b f ) Input gate: i t =σ(W i ·[h′ t-1 , x t ]+b i ) Candidate cell states: Cell status update: Output gate: o t =σ(W o ·[h′ t-1 , x t ]+b o ) Hidden state output: h′ t =o t ·tanh(c t ) Among them, W hh , W xh , W f , W i , W c and W o is the weight matrix, b h 、b f 、b i 、b c and b o is the bias term; Parameter initialization: The weight matrix W in the model hh , W zh , W f , W i , W c and W o and the bias term b h , b f , b i , b c and b o Random initialization using normal or uniform distribution; Loss function: In order to train the model, a loss function is defined, and the calculation formula of the loss function is: Where n is the number of samples, y i is the actual value of the ith sample, is the predicted value of the ith sample.

5. A big data asset management system according to claim 3, characterized in that: The model design and training unit performs model training, including the following steps: Data preparation: Collect and integrate historical data within the asset management system and external data sources to form a training data set, which includes input sequence X and corresponding output sequence Y; Forward propagation: input the training data into the model, forward propagate through the RNN layer and LSTM layer, and get the predicted value Calculate loss: Use the loss function to calculate the predicted value The difference between the actual value Y and the loss value is obtained; Back propagation and parameter update: The gradient of the loss value to the model parameters is calculated through the back propagation algorithm, and the model parameters are updated using Adam; Iterative training: Repeat the process of forward propagation, loss calculation, and backpropagation and parameter update until the preset number of iterations is reached or the loss value converges.

6. A big data asset management system according to claim 1, characterized in that: The credit assessment module includes a data integration unit, a credit score calculation unit, a high-risk user identification unit and a credit management unit. The data integration unit is responsible for acquiring data from various data sources, and performing cleaning, conversion and integration, and passing the integrated data to the credit score calculation unit and the high-risk user identification unit. The credit score calculation unit calculates the credit score of the user. The credit score calculation unit receives the integrated data transmitted by the data integration unit, generates a credit score after calculation, and passes the score result to the high-risk user identification unit and the credit management unit. The high-risk user identification unit receives the credit score transmitted by the credit score calculation unit, generates a high-risk user list after risk analysis, and passes the result to the credit management unit. The credit management unit takes corresponding management measures according to the user's credit score and risk level. The credit management unit receives the high-risk user list transmitted by the high-risk user identification unit and the credit score transmitted by the credit score calculation unit, and generates corresponding management instructions after making a decision.

7. A big data asset management system according to claim 2, characterized in that: The specific steps of credit score calculation by the credit score calculation unit are as follows: Data preprocessing: cleaning and organizing raw data; Feature selection: select features that have an impact on credit scores; feature Scaling: Use standardization to scale so that different features have the same weight in the model. The standardization formula is Among them, x is the original data, μ is the mean, and σ is the standard deviation; Model building: Building a credit scoring model based on selected features and logistic regression; Model evaluation: Use test set data for verification and evaluate the accuracy and reliability of the model; Credit score calculation: The user's feature data is input into the credit score model to obtain the prediction result, which is then converted into a credit score based on the score mapping table.

8. A big data asset management system according to claim 7, characterized in that: The model is constructed in the following specific steps: Data preparation: Collect and organize the dataset for model training, including the selected features and corresponding credit scores; Feature engineering: Use standardization to scale features so that they have equal weight in the model, and select or remove features as needed; Model training: Use the data processed by logistic regression for training to obtain the credit scoring model. The logistic regression prediction formula is: Where Y is the target variable, X is the feature vector, β0, β1, ..., β n are the parameters of the model. Initialize the model parameters, select the logarithmic loss function and the gradient descent method, iteratively update the model parameters until the convergence condition is reached, and find a set of optimal parameters β so that the model has the highest prediction accuracy on the training set; Parameter estimation: After training is completed, the estimated value of the model parameter β is obtained, which reflects the influence of each feature on the target variable Y; Model evaluation: Use test set data for verification; Model deployment: Integrate the model into the credit scoring calculation unit.

Images