Remanufactured product recovery prediction method and system based on artificial intelligence model

Abstract

The invention belongs to the technical field of machine learning, and discloses a remanufactured product recovery prediction method based on an artificial intelligence model, and the method carries out the prediction of the recovery amount of a remanufactured product through employing a plurality of artificial intelligence models (such as a support vector machine, a random forest, a decision tree, LSTM, TCN, Transformer, N-BEATS and the like), and remarkably improves the accuracy and stability of recovery prediction. Compared with a traditional method, the method has the advantages that nonlinear relation and complex time sequence data can be better processed, and the limitation of a traditional model in processing long-time dependence and high-dimensional data is overcome. Besides, through a multi-model integration and evaluation mechanism, a decision maker can select the most suitable prediction model according to different situations, the recovery decision is optimized, the robustness and the overall prediction capability of the system are improved, and the method is obviously superior to the existing traditional recovery prediction technology.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning, and particularly relates to a remanufactured product recovery prediction method and system based on an artificial intelligence model. Background Technique

[0002] The recovery of remanufacturing reverse logistics is restricted by its own characteristics. For example, the quality of waste products is unknown, the recovery mechanism of waste products is unknown, etc. All these will bring certain troubles to the research of remanufacturing reverse logistics recovery and hinder the development of remanufacturing reverse logistics recovery. The existing remanufactured product recovery methods mainly include time series analysis, regression analysis, statistical analysis, and artificial intelligence models.

[0003] Time series prediction involves predicting future data by analyzing and modeling historical data, which has been widely applied in various fields. L.S. Vivek et al. used a time series model to estimate how many mobile phones will be recycled or discarded in India in 2020. The Holt trend exponential smoothing and autoregressive integrated moving average (ARIMA) models were applied, and the most suitable model was selected based on the trends and seasonality in the data. Neha Kapadia et al. proposed a time series-based energy-saving recovery prediction model to predict the recovery of remanufactured products. For the time series model, the ARIMA model has the best prediction effect, but the ARIMA model performs poorly in dealing with non-linear relationships. At the same time, the data of the ARIMA must be smooth. Especially when dealing with long time series data, its performance will decrease significantly.

[0004] As a commonly used method in statistics, regression analysis is also often applied to predictions in many fields. Tasneem Imtiyaz Zargar [4] took Rajouri as an example. The research area was divided into different sample points according to the local population density, and municipal solid waste on weekdays, weekends, and festivals was collected at four locations at each point to predict the trends of waste generation and accumulation. Li Peng et al. used linear and non-linear regression algorithms to provide the accuracy of prediction and identification trends through a linear regression model and the non-linear regression model in terms of prediction results, and predicted the recycling of remanufactured products. Runqi Wang used the support vector regression (SVR) algorithm to construct an ash fusion temperature (AFT) model for the bottom ash of municipal solid waste incinerators (MSWIs) to predict the AFT of the same type of MSWI bottom ash, including fly ash, bottom slag, and other ash. In addition to regression analysis, some other statistical methods also play a crucial role in the recycling management of solid waste. Zheng Xuan Hoy used the method of integrated uncertainty analysis to solve the problem of insufficient data in the prediction of urban solid waste generation trends. This study focused on the case of Malaysia as an emerging economy. Through this model, it was possible to predict the physical composition trends of urban solid waste at the national scale, including the generation amounts of various waste components. However, these above regression models assume linear relationships, feature engineering relies on expert knowledge, has weak capabilities in dealing with complex data structures and high-dimensional data, limited generalization and flexibility, and poor scalability. In contrast, artificial intelligence models such as deep learning can automatically extract features, adapt to non-linear relationships and large-scale data, and have stronger generalization capabilities.

[0005] In recent years, society and humanity have become increasingly dependent on artificial intelligence technology, which is an important symbol of the prosperity of human society. Artificial intelligence (AI) algorithms play a crucial role in predicting the recycling of remanufactured products. By processing and analyzing large amounts of data, they provide efficient and accurate recycling predictions. Wenjing Lu et al. used an advanced machine learning algorithm, the gradient boost regression tree, to develop a recycling prediction model (WGMod), and selected annual precipitation, population density, and annual average temperature as key influencing factors to predict the generation of remanufactured products. Zhenying Zhang

[30] et al. used the total urban population, total retail sales of consumer goods, per capita urban consumption expenditure, tourism, etc. as model inputs, and predicted the generation of municipal solid waste through a combination of hybrid learning models. By using relevant features as model inputs and training and predicting the model based on some historical data to achieve the target results, the rise and maturity of AI methods have enabled a deeper understanding of various fields and more accurate predictions. Artificial intelligence models are widely praised for their advantages, including time savings, high prediction accuracy when applied to complex non-linear problems, and significant reduction in human and resource consumption in unnecessary repeated experiments. In addition, since the amount of data in the field of remanufactured product recycling is usually not too large, the computational cost of artificial intelligence methods during the training process is much lower than the potential benefits they can bring.

[0006] Through the above analysis, the problems and defects existing in the prior art are as follows:

[0007] (1) For time series models, the ARIMA model has the best prediction effect, but the ARIMA model performs poorly in dealing with non-linear relationships. At the same time, the data of the ARIMA model must be smooth. Especially when dealing with long time series data, its performance will decrease significantly.

[0008] (2) Statistical regression models assume linear relationships, feature engineering relies on expert knowledge, has weak capabilities in dealing with complex data structures and high-dimensional data, limited generalization and flexibility, and poor scalability. Summary of the Invention

[0009] In view of the problems existing in the prior art, the present invention provides a method for predicting the recycling of remanufactured products based on an artificial intelligence model.

[0010] The present invention is implemented as follows. A method for predicting the recycling of remanufactured products based on an artificial intelligence model includes:

[0011] Step 1, building a model for predicting the recycling of remanufactured products:

[0012] First, define the key factors for the recycling of remanufactured products, and generate the waste recycling status by establishing a recycling volume prediction model; construct a remanufactured product recycling prediction model that includes factors such as daily gross production value and daily commodity retail number, with the main purpose of studying the impact of these factors on the recycling of remanufactured products;

[0013] Step 2, in this experiment alone, design an artificial intelligence model based on recycling volume prediction, and use a variety of machine learning algorithms, including Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), BP Neural Network (BP), Long Short-Term Memory Network (LSTM), Temporal Convolutional Network (TCN), Transformer, and N-BEATS;

[0014] Step 3, design the evaluation metrics for the algorithms:

[0015] Design the evaluation metrics for the algorithms to intuitively reflect the fitting degree of the algorithms to the data and the accuracy of predicting the target motion elements;

[0016] Step 4, evaluate the advantages, disadvantages, and applicable ranges of each model by comparing and analyzing the performance of these models on different recycling station datasets;

[0017] Step 5, visualization of algorithm prediction:

[0018] Use matplotlib to achieve the visualization of algorithm prediction, and clearly display the comparison between the prediction results and the true values of each algorithm through charts;

[0019] Step 6, user interface design:

[0020] Create a user interface based on Tkinter in Python for the visual display of the recycling of remanufactured products; provide instant algorithm comparison and data display.

[0021] Furthermore, the prediction modeling of the recycling of remanufactured products:

[0022] First, for the prediction of the recycling of remanufactured products, define the key factors for recycling volume prediction, including economic factors such as daily gross production value and daily commodity retail number; these factors are considered to have a significant impact on the recycling volume of remanufactured products, so they are used as the input features of the model; specifically, explore the impact of these factors on the recycling volume of remanufactured products, and use the following evaluation metrics:

[0023] Use the Pearson correlation coefficient to explore the correlation between these factors and the recycling of remanufactured products. In addition, combine the p-value and t-value to evaluate the correlation; the definitions of each metric are as follows:

[0024] (1) The Pearson correlation coefficient is used to measure the linear correlation between two variables, and its value range is from -1 to 1; the specific formula is:

[0025]

[0026] where r is the Pearson correlation coefficient, xi and yi are the observed values of variables X and Y, and are the means of X and Y;

[0027] (2) The t-value is used to test whether the sample data significantly deviates from the null hypothesis and is usually used for the coefficient significance test in regression analysis; the specific formula is:

[0028]

[0029] where t is the t-value, r is the Pearson correlation coefficient, and n is the total number of samples;

[0030] (3) The p-value is used to evaluate the probability of the observed result occurring under the null hypothesis and is an indicator for testing the significance of correlation.

[0031] Furthermore, the design is based on an artificial intelligence model for recycling volume prediction:

[0032] Multiple machine learning algorithms are adopted, including Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), BP Neural Network (BP), Long Short-Term Memory Network (LSTM), Temporal Convolutional Network (TCN), Transformer, and N-BEATS;

[0033] (1) Support Vector Machine (SVM)

[0034] SVM is a typical classification and regression algorithm suitable for handling prediction tasks of small samples and non-linear data; in this experiment, SVM is used for the prediction of recycling volume, and different types of input data are processed by selecting appropriate kernel functions such as linear kernel and radial basis kernel; the parameters of SVM are optimized through cross-validation and grid search to obtain the best model;

[0035] (2) Random Forest (RF)

[0036] Random Forest is an ensemble learning method that improves the accuracy and stability of prediction by integrating multiple decision trees; in the prediction of solid waste recycling, RF can effectively handle high-dimensional data and has strong anti-overfitting ability; by training multiple decision trees, Random Forest can capture the complex non-linear relationship between recycling volume and economic indicators;

[0037] (3) Decision Tree (DT)

[0038] Decision tree is a common supervised learning algorithm that makes predictions by recursively dividing data into different subsets; in the task of predicting the recycling volume, the decision tree model can better capture the relationship between specific economic factors and the recycling volume;

[0039] (4) BP neural network BP

[0040] BP neural network is a common feedforward neural network that is trained through the error backpropagation algorithm; in the prediction of solid waste recycling volume, the BP network can effectively fit complex non-linear relationships and perform feature abstraction through a hierarchical structure; by continuously adjusting the number of network layers and nodes, the prediction ability of the model is optimized;

[0041] (5) Long short-term memory network LSTM

[0042] LSTM is a recurrent neural network RNN that is particularly suitable for processing time series data; the change of solid waste recycling volume has time series characteristics, so LSTM can effectively capture the law of recycling volume changing with time through its memory unit; by training the LSTM network, the future change of recycling volume can be predicted within a long time range, which is especially suitable for capturing the lag effect of economic factors;

[0043] (6) Temporal convolutional network TCN

[0044] TCN is a method for processing time series data based on the convolutional neural network CNN; different from LSTM, TCN captures the time series pattern of data through one-dimensional convolutional operations, and has higher parallel processing ability and longer receptive field.

[0045] Furthermore, the Transformer:

[0046] Transformer is a deep learning model based on the self-attention mechanism, which is widely used in natural language processing tasks; in the prediction of recycling volume, Transformer can effectively model the global dependencies between data, especially having advantages on large-scale datasets; through the multi-head self-attention mechanism, Transformer can capture the complex associations between the recycling volume and various economic factors.

[0047] Furthermore, the N-BEATS:

[0048] N-BEATS is a deep learning model suitable for time series prediction, and is particularly good at capturing long-term dependencies.

[0049] Furthermore, the design of the evaluation metrics of the algorithm:

[0050] The model uses three evaluation metrics to evaluate the model performance: Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-squared (R2). The definitions of each metric are as follows:

[0051] (1) Mean Squared Error (MSE) is a metric that measures the difference between the predicted value and the true value. The smaller its value, the better the performance of the prediction model. The specific formula is as follows:

[0052]

[0053] n is the number of samples, y i is the corresponding true value, is the predicted value;

[0054] (2) Root Mean Squared Error (RMSE) is the square root of the Mean Squared Error. It is used to measure the difference between the predicted value and the true value, just like MSE. RMSE has the same unit as the original data and is easier to interpret. The specific formula is as follows:

[0055]

[0056] (3) R-squared (R2) represents the degree of data interpretation by the model, and its value ranges from 0 to 1. The closer the value is to 1, the stronger the model's ability to interpret the data;

[0057]

[0058] Another object of the present invention is to provide a remanufactured product recovery prediction system based on an artificial intelligence model, including:

[0059] A prediction modeling module for remanufactured product recovery prediction modeling: First, define the key factors for remanufactured product recovery, and generate the waste recovery status by establishing a recovery quantity prediction model; construct a remanufactured product recovery prediction model including factors such as daily gross production value and daily commodity retail number, mainly aiming to study the impact of these factors on remanufactured product recovery; algorithms. In this experiment, design an artificial intelligence model based on recovery quantity prediction, and adopt a variety of machine learning algorithms, including Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), BP Neural Network (BP), Long Short-Term Memory Network (LSTM), Temporal Convolutional Network (TCN), Transformer, and N-BEATS;

[0060] An index design module for the evaluation index design of the algorithm: Design evaluation indexes for the algorithm to intuitively reflect the fitting degree of the algorithm to the data and the accuracy of the prediction of the target motion elements;

[0061] A comparative analysis module for evaluating the advantages, disadvantages, and applicable ranges of each model by comparing and analyzing the performance of these models on different recycling station datasets;

[0062] A visualization module for visualizing algorithm predictions: It uses matplotlib to implement the visualization of algorithm predictions, and clearly shows the comparison between the prediction results and the true values of each algorithm through charts.

[0063] An interface design module for creating a user interface based on Tkinter in Python for visual display of remanufactured product recycling; providing instant algorithm comparison and data display.

[0064] Another object of the present invention is to provide a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of the remanufactured product recycling prediction method based on the artificial intelligence model.

[0065] Another object of the present invention is to provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the remanufactured product recycling prediction method based on the artificial intelligence model.

[0066] Another object of the present invention is to provide an information data processing terminal for implementing the remanufactured product recycling prediction system based on the artificial intelligence model.

[0067] Combined with the above technical solutions and solved technical problems, please analyze the advantages and positive effects of the technical solution to be protected by the present invention from the following aspects:

[0068] First, aiming at the technical problems existing in the above-mentioned prior art and the difficulty of solving this problem, closely combining the technical solution to be protected by the present invention and the results and data in the R & D process, etc., analyze in detail and deeply how the technical solution of the present invention solves the technical problems and the creative technical effects brought after solving the problems. The specific description is as follows:

[0069] The object of the present invention is to apply advanced artificial intelligence to the prediction of remanufactured product recycling. Aiming at the problems that the past technologies perform poorly in the face of non-linear relationships and complex situations, methods such as machine learning and deep learning are used to improve the recycling prediction ability of remanufactured products and reduce the computational complexity, and at the same time help relevant decision-makers to select appropriate models in the face of different remanufactured product recycling scenarios.

[0070] The present invention aims to solve the problems that in the face of a complex and uncertain environment in the recycling prediction of remanufactured products, including poor handling of non-linear relationships, strong dependence on the initial state, and poor model stability, resulting in difficulties in meeting the actual requirements of the recycling prediction effect and system robustness of remanufactured products. At the same time, through the evaluation of the effects of artificial intelligence models, the present invention can also help decision-makers select appropriate models in different recycling scenarios of remanufactured products.

[0071] To solve these problems, the present invention proposes a recycling prediction technology for remanufactured products based on artificial intelligence models. Multiple models (support vector machine, random forest, decision tree, BP neural network, long short-term memory network, temporal convolutional network, transformer, N-BEATS) are used to predict the recycling of remanufactured products, and the relevant influencing factors of product recycling prediction are considered to improve the prediction ability. At the same time, through the comparison of the prediction effects of multiple models, the application scenarios of these models in the recycling prediction of remanufactured products are summarized, so as to help decision-makers select the most suitable model in different scenarios.

[0072] The experimental results show that the artificial intelligence model has stronger stability in terms of prediction accuracy, computational efficiency, and overall system robustness in remanufactured products, can better meet various actual application requirements, and is significantly superior to existing traditional technologies.

[0073] 1. Recycling prediction technology for remanufactured products based on artificial intelligence models. The core innovation of the present invention lies in proposing a recycling prediction technology for remanufactured products based on artificial intelligence models, aiming to solve the limitations of traditional recycling prediction methods (such as ARIMA, regression analysis, etc.) in the face of problems such as non-linear relationships, complex time series data, and data uncertainty. This technology accurately predicts the recycling volume of remanufactured products through multiple advanced machine learning and deep learning models (such as support vector machine, random forest, decision tree, BP neural network, LSTM, TCN, Transformer, N-BEATS, etc.), and combines the advantages and disadvantages of different models to help decision-makers select the most suitable model for application.

[0074] 2. Multi-model integration and its evaluation mechanism: The present invention integrates multiple machine learning algorithms (such as support vector machine, random forest, decision tree, etc.) and deep learning algorithms (such as LSTM, TCN, Transformer, N-BEATS, etc.) to solve the problem of inconsistent performance of a single algorithm in different scenarios. By comparing the prediction effects of these models, evaluation metrics (such as RMSE, MSE, R 2Quantify the performance of each model (etc.), further helping decision-makers select the most suitable model, improving the accuracy of prediction and the robustness of the system. Implementation and application of ensemble learning: Combine the least squares method and the ELM algorithm, utilize the advantages of both to achieve ensemble learning, so as to adapt to different target motion scenarios and improve the accuracy and stability of target motion state estimation.

[0075] 3. Dynamic recovery quantity prediction modeling and correlation analysis of economic factors: By establishing a dynamic model based on recovery quantity prediction, the present invention considers the influence of economic factors such as daily gross production value and daily commodity retail volume on the recovery quantity of remanufactured products. By using statistical methods such as Pearson correlation coefficient, t-value, and p-value, accurately evaluate the influence degree of these factors on the recovery quantity, and incorporate these factors into the artificial intelligence model, thereby improving the accuracy and adaptability of the recovery prediction model.

[0076] 4. Modeling ability for temporal and non-linear relationships: LSTM and other deep learning models (such as TCN, Transformer, etc.) can effectively capture the temporal characteristics and non-linear relationships of the recovery quantity, which are difficult for traditional regression models to handle. LSTM is particularly suitable for processing data with long-term dependencies, and can discover patterns in the time variation of the recovery quantity, thereby achieving accurate recovery prediction.

[0077] 5. User interface design and visualization display: The present invention also provides a user interface based on Python Tkinter for visualizing the recovery prediction of remanufactured products. Users can intuitively select different recovery quantity prediction models through the interface and view the imported historical data and prediction results. The interface supports real-time data display, algorithm comparison, and display of evaluation metrics (such as MSE, RMSE, R 2 etc.), improving the user's interaction experience and the convenience of algorithm tuning.

[0078] Technical application and decision support: The technology of the present invention not only improves the prediction ability of the recovery of remanufactured products, but also through multi-model comparison and evaluation, can provide more accurate and stable prediction results for decision-makers, helping them select the most suitable model for application in the face of different recovery situations. This is of great significance for improving resource recovery efficiency and enhancing the environmental and economic benefits of the remanufacturing industry.

[0079] The present invention predicts the recycling quantity of remanufactured products by using a variety of artificial intelligence models (such as support vector machines, random forests, decision trees, LSTM, TCN, Transformer, N-BEATS, etc.), significantly improving the accuracy and stability of recycling prediction. Compared with traditional methods, the present invention can better handle non-linear relationships and complex time-series data, overcoming the limitations of traditional models in dealing with long-term dependencies and high-dimensional data. In addition, through multi-model integration and evaluation mechanisms, decision-makers can select the most suitable prediction model according to different scenarios, optimize recycling decisions, improve the robustness of the system and the overall prediction ability, and significantly outperform existing traditional recycling prediction technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 FIG. is a flowchart of a method for predicting the recycling of remanufactured products based on an artificial intelligence model provided by an embodiment of the present invention.

[0081] Figure 2 FIG. is a block diagram of the structure of a system for predicting the recycling of remanufactured products based on an artificial intelligence model provided by an embodiment of the present invention.

[0082] Figure 3 FIG. is an SVM classification diagram in the kernel space provided by an embodiment of the present invention.

[0083] Figure 4 FIG. is an LSTM network structure diagram provided by an embodiment of the present invention.

[0084] Figure 5 FIG. is a flowchart of predicting the recycling of remanufactured products provided by an embodiment of the present invention.

[0085] Figure 6 FIG. is a comparison diagram of the prediction effects of the recycling of remanufactured products in different regions provided by an embodiment of the present invention.

[0086] Figure 7 FIG. is an interface diagram for predicting the recycling quantity provided by an embodiment of the present invention.

[0087] Figure 8 FIG. is an interface diagram for calling a prediction model for the recycling of remanufactured products provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0088] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0089] The remanufactured product recycling prediction method based on an artificial intelligence model takes the modeling of remanufactured product recycling prediction as the core starting point of the whole process. By analyzing the key factors affecting the recycling volume, such as the daily gross production value and the daily number of commodity retail sales, a recycling prediction model is constructed. The purpose of this model is to reveal the relationship between the recycling volume of remanufactured products and these external economic factors, providing basic support for the prediction of the waste recycling status. By collecting historical data, a preliminary mathematical description and correlation model are established, laying a data foundation for the training of subsequent artificial intelligence algorithms.

[0090] Based on the modeling, multiple artificial intelligence algorithms are designed to predict the recycling volume of remanufactured products, including Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), BP Neural Network (BP), Long Short-Term Memory Network (LSTM), Temporal Convolutional Network (TCN), Transformer, and N-BEATS, etc. These algorithms have their own focuses. Some are suitable for handling short-term prediction tasks (such as SVM and DT), some are suitable for time series analysis (such as LSTM and TCN), and some have strong adaptability to long-cycle data and complex causal relationships (such as Transformer and N-BEATS).

[0091] To intuitively reflect the performance of each algorithm, evaluation indicators including Mean Squared Error (MSE), Mean Absolute Error (MAE), and R 2 are designed. These indicators are used to quantify the model's ability to fit data and the accuracy of the prediction target. By comprehensively using these evaluation indicators, the advantages and limitations of each algorithm in different scenarios can be effectively identified, providing a basis for selecting the optimal model.

[0092] For the historical data sets of different recycling stations, the performance of multiple models is systematically compared and analyzed, and their advantages, disadvantages, and applicable ranges are evaluated. For example, Random Forest (RF) performs well when the data feature diversity is high, while LSTM has significant advantages in capturing time series characteristics. Through this analysis process, the most suitable algorithm combination can be selected for a specific scenario, improving the overall prediction efficiency and accuracy.

[0093] To make the prediction results more intuitive and clear, Matplotlib is used to visualize the prediction results of the model. The predicted values and actual values of each model are plotted on the same chart, clearly showing the fitting effect and error distribution of the model. At the same time, through chart comparison, it is easy to quickly discover the abnormal prediction points or regular deviations of the model, assisting users in making judgments and decisions.

[0094] Finally, a user interface based on Python Tkinter is designed to provide a friendly interaction platform for users. This interface supports the instant display of the prediction results of different algorithms and their evaluation metrics, and at the same time supports users to customize the input data to view the prediction performance of the corresponding models. Through this visual and interactive design, users can quickly compare different models and select the prediction algorithm that best meets the actual needs, greatly improving the efficiency and convenience in practical applications.

[0095] Through scientific modeling, multi-algorithm fusion, visualization technology, and human-computer interaction design, this overall workflow realizes the efficiency, accuracy, and visualization of the remanufactured product recovery prediction.

[0096] As Figure 1 shown, a method for predicting the recovery of remanufactured products based on an artificial intelligence model provided by an embodiment of the present invention includes the following steps:

[0097] S101, Modeling for predicting the recovery of remanufactured products:

[0098] First, define the key factors for the recovery of remanufactured products, and generate the waste recovery status by establishing a recovery volume prediction model; construct a remanufactured product recovery prediction model that includes factors such as daily gross production value and daily commodity retail number, and the main purpose is to study the impact of these factors on the recovery of remanufactured products;

[0099] S102, In this experiment, design an artificial intelligence model based on the prediction of the recovery volume, and adopt a variety of machine learning algorithms, including support vector machine SVM, random forest RF, decision tree DT, BP neural network BP, long short-term memory network LSTM, temporal convolutional network TCN, Transformer, N-BEATS;

[0100] S103, Design of evaluation metrics for the algorithm:

[0101] Design evaluation metrics for the algorithm to intuitively reflect the fitting degree of the algorithm to the data and the accuracy of predicting the target motion elements;

[0102] S104, Evaluate the advantages, disadvantages, and applicable ranges of each model by comparing and analyzing the performance of these models on different recycling station datasets;

[0103] S105, Visualization of algorithm prediction:

[0104] Use matplotlib to realize the visualization of algorithm prediction, and clearly display the comparison between the prediction results of each algorithm and the true values through charts;

[0105] S106, User interface design:

[0106] Create a user interface based on Tkinter in Python for visual display of remanufactured product recycling; provide instant algorithm comparison and data display.

[0107] The remanufactured product recycling prediction modeling provided by the embodiments of the present invention:

[0108] First, for the prediction of remanufactured product recycling, define the key factors for predicting the recycling volume, including economic factors such as daily gross production value and daily commodity retail number; these factors are considered to have a significant impact on the recycling volume of remanufactured products, so they are used as input features of the model; specifically, explore the impact of these factors on the recycling volume of remanufactured products and use the following evaluation indicators:

[0109] Use the Pearson correlation coefficient to explore the correlation between these factors and the recycling of remanufactured products. In addition, combine the p-value and t-value to evaluate the correlation; the definitions of each indicator are as follows:

[0110] (1) The Pearson correlation coefficient is used to measure the linear correlation between two variables, and its value range is from -1 to 1; the specific formula is:

[0111]

[0112] where r is the Pearson correlation coefficient, xi and yi are the observed values of variables X and Y, and are the means of X and Y;

[0113] (2) The t-value is used to test whether the sample data significantly deviates from the null hypothesis and is usually used for the coefficient significance test in regression analysis; the specific formula is:

[0114]

[0115] where t is the t-value, r is the Pearson correlation coefficient, and n is the total number of samples;

[0116] (3) The p-value is used to evaluate the probability of the observed result occurring under the condition that the null hypothesis holds and is an index for testing the significance of the correlation.

[0117] The artificial intelligence model based on recycling volume prediction provided by the embodiments of the present invention:

[0118] Adopt a variety of machine learning algorithms, including Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), BP Neural Network (BP), Long Short-Term Memory Network (LSTM), Temporal Convolutional Network (TCN), Transformer, and N-BEATS;

[0119] (1) Support Vector Machine (SVM)

[0120] SVM is a typical classification and regression algorithm, suitable for handling prediction tasks of small samples and non-linear data; in this experiment, SVM is used for the prediction of the recovery amount, and appropriate kernel functions such as linear kernel and radial basis kernel are selected to process different types of input data; the parameters of SVM are optimized through cross-validation and grid search to obtain the best model;

[0121] (2) Random Forest RF

[0122] Random Forest is an ensemble learning method that improves the accuracy and stability of prediction by integrating multiple decision trees; in the prediction of solid waste recovery, RF can effectively handle high-dimensional data and has strong anti-overfitting ability; by training multiple decision trees, Random Forest can capture the complex non-linear relationship between the recovery amount and economic indicators;

[0123] (3) Decision Tree DT

[0124] Decision Tree is a common supervised learning algorithm that makes predictions by recursively dividing data into different subsets; in the recovery amount prediction task, the decision tree model can better capture the relationship between specific economic factors and the recovery amount;

[0125] (4) BP Neural Network BP

[0126] BP Neural Network is a common feedforward neural network that is trained by the error backpropagation algorithm; in the prediction of solid waste recovery amount, the BP network can effectively fit complex non-linear relationships and perform feature abstraction through a hierarchical structure; by continuously adjusting the number of network layers and nodes, the prediction ability of the model is optimized;

[0127] (5) Long Short-Term Memory Network LSTM

[0128] LSTM is a recurrent neural network RNN that is particularly suitable for processing time series data; the change of solid waste recovery amount has time series characteristics, so LSTM can effectively capture the law of the recovery amount changing over time through its memory unit; by training the LSTM network, the future change of the recovery amount can be predicted within a long time range, which is particularly suitable for capturing the lag effect of economic factors;

[0129] (6) Temporal Convolutional Network TCN

[0130] TCN is a method for processing time series data based on convolutional neural network CNN; different from LSTM, TCN captures the time series pattern of data through one-dimensional convolution operation, and has higher parallel processing ability and longer receptive field.

[0131] The Transformer provided by the embodiments of the present invention:

[0132] Transformer is a deep learning model based on self-attention mechanism and is widely used in natural language processing tasks; in recycling volume prediction, Transformer can effectively model the global dependencies between data and has advantages especially on large-scale datasets; through the multi-head self-attention mechanism, Transformer can capture the complex associations between recycling volume and various economic factors.

[0133] N-BEATS provided by the embodiments of the present invention:

[0134] N-BEATS is a deep learning model suitable for time series prediction and is especially good at capturing long-term dependencies.

[0135] Evaluation index design of the algorithm provided by the embodiments of the present invention:

[0136] The model uses three evaluation indexes to evaluate the model effect: mean squared error (MSE), root mean squared error (RMSE), and correlation coefficient (R2). The definitions of each index are as follows:

[0137] (1) Mean squared error (MSE) is an index to measure the difference between the predicted value and the true value. The smaller its value is, the better the performance of the prediction model is. The specific formula is as follows:

[0138]

[0139] n is the number of samples, y i is the corresponding true value, is the predicted value;

[0140] (2) Root mean squared error (RMSE) is the square root of the mean squared error. It is used to measure the difference between the predicted value and the true value just like the mean squared error. RMSE has the same unit as the original data and is easier to interpret. The specific formula is as follows:

[0141]

[0142] (3) Correlation coefficient R2 represents the degree of explanation of the model to the data, and its value range is between 0 and 1. The closer the value is to 1, the stronger the ability of the model to explain the data;

[0143]

[0144] As Figure 2 shown, a remanufactured product recycling prediction system based on an artificial intelligence model provided by the embodiments of the present invention includes:

[0145] Prediction Modeling Module, used for predicting and modeling the remanufactured product recycling: First, define the key factors for remanufactured product recycling, and generate the waste recycling status by establishing a recycling volume prediction model; construct a remanufactured product recycling prediction model including factors such as daily gross production value and daily commodity retail number, with the main purpose of studying the impact of these factors on remanufactured product recycling; Algorithm, in this experiment, design an artificial intelligence model based on recycling volume prediction, and adopt a variety of machine learning algorithms, including Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), BP Neural Network (BP), Long Short-Term Memory Network (LSTM), Temporal Convolutional Network (TCN), Transformer, N-BEATS;

[0146] Index Design Module, used for designing evaluation indexes of the algorithm: Design evaluation indexes for the algorithm to intuitively reflect the fitting degree of the algorithm to the data and the accuracy of predicting the target motion elements;

[0147] Comparison and Analysis Module, used for evaluating the advantages, disadvantages and applicable scopes of each model by comparing and analyzing the performances of these models on different recycling station data sets;

[0148] Visualization Module, used for visualizing the algorithm prediction: Use matplotlib to realize the visualization of the algorithm prediction, and clearly display the comparison between the prediction results and the true values of each algorithm through charts;

[0149] Interface Design Module, used for creating a user interface based on Tkinter in Python for visual display of remanufactured product recycling; Provide instant algorithm comparison and data display.

[0150] Another object of the present invention is to provide a computer device, the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the remanufactured product recycling prediction method based on the artificial intelligence model.

[0151] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the processor executes the steps of the remanufactured product recycling prediction method based on the artificial intelligence model.

[0152] Another object of the present invention is to provide an information data processing terminal, and the information data processing terminal is used to implement the remanufactured product recycling prediction system based on the artificial intelligence model.

[0153] Specific implementation of the present invention:

[0154] The models used in this invention include Support Vector Machine, Random Forest, Decision Tree, BP Neural Network, Long Short-Term Memory Network, Transformer, and N-BEATS model. Decision Tree (DT for short) is a machine learning algorithm for classification and regression tasks. It constructs a tree structure by recursively splitting the dataset into smaller subsets. Each node represents a feature, each branch represents a value condition, and the leaf node represents the decision result. Its advantages include being easy to understand and interpret, not requiring data preprocessing, being able to handle multiple types of data, and having a feature selection function. However, its disadvantages are being prone to overfitting, unstable, favoring multi-valued attributes, and having a relatively high computational complexity. When analyzing data, the Decision Tree model performs data splitting based on the decision rules of input features, providing a non-parametric method to identify complex patterns. By using the Decision Tree model, a prediction tool is provided for the recycling of remanufactured products

[11] . Support Vector Machine (SVM for short) is a supervised learning model used for classification and regression analysis. It separates different classes in the dataset by finding a hyperplane that maximizes the class interval. SVM performs excellently in both linear and non-linear classification tasks. For non-linear problems, SVM uses kernel functions to map the data into a higher-dimensional feature space to achieve linear separation. The SVM optimization problem can be written in the following form:

[0155]

[0156] where C is the number of samples, max(0, 1 - t i (Φ T (x i )w + w 0 )) is the hinge loss function, Φ is the kernel mapping, and w = (w 0 , w 1 ,..., w N ) are the unknown weights to be estimated. The classification graph of SVM in the kernel space is shown in Figure 3 . Research

[12] used Support Vector Machine (SVM) to solve the recycling prediction problem of remanufactured products. The main advantage is its ability to adapt to the changing characteristics of data, especially performing well when there is less training data. By using kernel functions (such as the radial basis function), SVM can achieve non-linear mapping of data, thus more accurately predicting the amount of waste generated. However, the disadvantages of SVM are also relatively obvious, manifested in its sensitivity to outliers and the weak interpretability of the high-dimensional mapping of kernel functions. When the amount of data increases, the performance of SVM is often mediocre.

[0157] Random Forest (RF) is an ensemble learning method that improves the prediction performance of classification or regression tasks by constructing multiple decision trees and combining their results. It has the advantages of high accuracy, anti-overfitting, ability to handle high-dimensional data, and strong robustness. However, it has large computational overhead, poor interpretability, high memory consumption, and complex parameter tuning. Random Forest performs well in cases with many features and high data noise and is suitable for a variety of practical applications. In the recycling of remanufactured products, Random Forest helps to maximize the environmental benefits and optimize the economic benefits of waste management by predicting product recovery volume, optimizing processing technologies, improving resource utilization efficiency, and providing data-driven decision support

[13] .

[0158] The BP neural network, namely the Backpropagation (BP) neural network, is a commonly used multi-layer feedforward neural network that trains the network weights through the error backpropagation algorithm. It mainly consists of an input layer, a hidden layer, and an output layer, can handle complex non-linear relationships, and has a wide range of applications in prediction problems in many fields. In solid waste management, the BP neural network can accurately predict solid waste through its powerful non-linear mapping, self-learning, and adaptive capabilities. The BP neural network improves the operation efficiency and prediction accuracy of the model and helps to achieve safe, efficient, and green coal mining and utilization of remanufactured products

[14] .

[0159] The Long Short-Term Memory (LSTM) network is a special type of Recurrent Neural Network (RNN), specifically designed to address the limitations of ordinary RNNs in dealing with long-term dependencies. LSTM controls the flow of information by introducing memory cells and gating mechanisms (input gate, forget gate, output gate), thus effectively capturing dependencies over long time spans.

[0160] The basic structure of LSTM includes a cell state and three gates: an input gate, a forget gate, and an output gate. The following are the calculation steps of LSTM at time step t.

[0161] (1) Forget gate:

[0162] f t = σ(W f ·[h t-1 , x t +b f ) (2)

[0163] where f represents the forget gate, σ is the activation function, W f is the weight matrix, [h t-1 , x t is the vector concatenation operation, bf is the bias vector; the forget gate controls which information in the cell state needs to be forgotten.

[0164] (2) Input gate:

[0165]

[0166] The input gate determines which new information needs to be added to the cell state, and the candidate memory unit state generates new candidate information.

[0167] (3) Update the cell state:

[0168]

[0169] The cell state combines the state at the previous time step and the new candidate information.

[0170] (4) Output gate and hidden state:

[0171]

[0172] In the aspect of remanufactured product recovery prediction, by introducing memory units and gating mechanisms, LSTM effectively solves the limitations of ordinary RNN in the long-term dependence problem. The LSTM (Long Short-Term Memory Network) model is widely used in the prediction of time series data

[18] . The structure diagram of LSTM is as Figure 4 shown. Among them, Xt represents the input data at time step t, ht represents the hidden state at time step t, A represents the three gating mechanisms, σ represents the sigmoid activation function, and tanH represents the hyperbolic tangent function.

[0173] The Temporal Convolutional Network (TCN for short) is a deep learning architecture for time series data modeling. It captures long-term dependence relationships through dilated convolution and causal convolution, and at the same time improves stability and training efficiency through residual connections. For this reason,

[43] proposed residual connections to improve the performance of very deep architectures, including adding the input of the TCN block to its output. And attempts are made to apply this model to the field of remanufactured product recovery.

[0174] The Transformer is a deep learning model based on the attention mechanism, capable of efficiently capturing long-range dependencies through multi-head self-attention mechanisms and parallel processing, and is widely used in tasks such as natural language processing and time series prediction. Its main advantages include efficient parallel computing and good long-range dependency modeling capabilities, while the main disadvantages are high computational and memory requirements, especially when dealing with very long sequences. Reference

[44] developed a new time series prediction method based on the Transformer architecture. The working principle of this method is to use the self-attention mechanism to learn complex and dynamic patterns from time series data. In our research, an implementation based on the studied Transformer architecture was used.

[0175] The present invention proposes a remanufactured product recovery prediction technology based on an artificial intelligence model, and its technical solution includes the following steps:

[0176] (1) Remanufactured product recovery prediction modeling: First, the key factors for remanufactured product recovery are defined, and a waste recovery status is generated by establishing a recovery volume prediction model. A remanufactured product recovery prediction model including factors such as daily gross production value and daily commodity retail volume is constructed, and the main purpose is to study the impact of these factors on the recovery of remanufactured products.

[0177] (2) In this experiment alone, an artificial intelligence model based on recovery volume prediction was designed, and a variety of machine learning algorithms were used, including support vector machine (SVM), random forest (RF), decision tree (DT), BP neural network (BP), long short-term memory network (LSTM), temporal convolutional network (TCN), Transformer, N-BEATS, etc.

[0178] (3) Design of evaluation metrics for the algorithm: Design evaluation metrics for the algorithm to intuitively reflect the fitting degree of the algorithm to the data and the accuracy of predicting the target motion elements.

[0179] (4) By comparing and analyzing the performance of these models on different recycling station datasets, the advantages, disadvantages, and applicable ranges of each model were evaluated.

[0180] (5) Visualization of algorithm prediction: Use matplotlib to achieve the visualization of algorithm prediction, and clearly display the comparison between the prediction results of each algorithm and the true values through charts.

[0181] (6) User interface design: Create a user interface based on Tkinter in Python for the visual display of remanufactured product recovery. Provide instant algorithm comparison and data display to enhance the user interaction experience and the intuitiveness of algorithm tuning.

[0182] The present invention will be described in detail below in conjunction with the accompanying drawings,Figure 5 It is the process framework diagram of the present invention.

[0183] 1. Remanufactured product recovery prediction modeling

[0184] First, for the prediction of remanufactured product recovery, this experiment defines the key factors for recovery volume prediction, including economic factors such as daily gross production value and daily commodity retail number. These factors are considered to have a significant impact on the recovery volume of remanufactured products, so they are used as input features of the model. Specifically, to explore the influence of these factors on the recovery volume of remanufactured products, the following evaluation indicators are used:

[0185] The Pearson correlation coefficient is used to explore the correlation between these factors and the recovery of remanufactured products. In addition, the p-value and t-value are combined to evaluate the correlation; the definitions of each indicator are as follows:

[0186] (1) The Pearson correlation coefficient is used to measure the linear correlation between two variables, and its value range is from -1 to 1. The specific formula is:

[0187]

[0188] where r is the Pearson correlation coefficient, xi and yi are the observed values of variables X and Y, and are the means of X and Y.

[0189] (2) The t-value is used to test whether the sample data significantly deviates from the null hypothesis, and is usually used for the coefficient significance test in regression analysis. The specific formula is:

[0190]

[0191] where t is the t-value, r is the Pearson correlation coefficient, and n is the total number of samples.

[0192] (3) The p-value is used to evaluate the probability of the observed result occurring under the condition that the null hypothesis holds, and is an indicator for testing the significance of the correlation.

[0193] To study the influence of these factors on the recovery volume of remanufactured products, a recovery volume prediction model is constructed. The core of the model is to predict based on the correlation between the time series characteristics of the recovery volume and economic factors. By reviewing historical recovery data and economic indicators, we find that there is a strong linear relationship between the recovery volume and the daily gross production value and the daily commodity retail number. Therefore, we use the artificial intelligence model introduced in the present invention to fit the relationship between these factors and the recovery volume, forming a preliminary recovery volume prediction model.

[0194] 2. Design of remanufactured product recovery prediction based on artificial intelligence model

[0195] In this experiment, to improve the accuracy of recovery quantity prediction, we adopted a variety of machine learning algorithms, including Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), BP Neural Network (BP), Long Short-Term Memory Network (LSTM), Temporal Convolutional Network (TCN), Transformer, and N-BEATS, etc. Each algorithm has different performances on the dataset, so a series of comparative analyses were conducted in this experiment.

[0196] (1) Support Vector Machine (SVM)

[0197] SVM is a typical classification and regression algorithm, suitable for handling prediction tasks of small samples and non-linear data. In this experiment, SVM was used for the prediction of recovery quantity, and appropriate kernel functions (such as linear kernel, radial basis kernel, etc.) were selected to process different types of input data. The parameters of SVM were optimized through cross-validation and grid search to obtain the best model.

[0198] (2) Random Forest (RF)

[0199] Random Forest is an ensemble learning method that improves the accuracy and stability of prediction by integrating multiple decision trees. In the prediction of solid waste recovery, RF can effectively handle high-dimensional data and has strong anti-overfitting ability. By training multiple decision trees, Random Forest can capture the complex non-linear relationship between recovery quantity and economic indicators, thus providing more accurate recovery quantity prediction.

[0200] (3) Decision Tree (DT)

[0201] Decision Tree is a common supervised learning algorithm that makes predictions by recursively dividing data into different subsets. In the recovery quantity prediction task, the decision tree model can better capture the relationship between specific economic factors and recovery quantity. Especially in the case of a small feature space or a large amount of missing data, the decision tree shows good interpretability and prediction effect.

[0202] (4) BP Neural Network (BP)

[0203] BP Neural Network is a common feedforward neural network that is trained through the error backpropagation algorithm. In the prediction of solid waste recovery quantity, the BP network can effectively fit complex non-linear relationships and perform feature abstraction through a hierarchical structure. We optimized the prediction ability of the model by continuously adjusting the number of network layers and nodes.

[0204] (5) Long Short-Term Memory Network (LSTM)

[0205] LSTM is a type of recurrent neural network (RNN) that is particularly suitable for processing time series data. The changes in the amount of solid waste recycled have a temporal sequence, so LSTM can effectively capture the pattern of the recycled amount changing over time through its memory cells. By training the LSTM network, we can predict the future changes in the recycled amount over a long time range, which is especially suitable for capturing the lag effect of economic factors.

[0206] (6) Temporal Convolutional Network (TCN)

[0207] TCN is a method for processing time series data based on Convolutional Neural Network (CNN). Different from LSTM, TCN captures the temporal pattern of data through one-dimensional convolutional operations, with higher parallel processing ability and a longer receptive field. In this experiment, TCN is used to handle long-term dependencies. Especially when the dataset is large, TCN can effectively improve the computational efficiency and prediction accuracy.

[0208] (7) Transformer

[0209] Transformer is a deep learning model based on the self-attention mechanism, which is widely used in natural language processing tasks. In the prediction of the recycled amount, Transformer can effectively model the global dependencies between data, especially having advantages on large-scale datasets. Through the multi-head self-attention mechanism, Transformer can capture the complex associations between the recycled amount and various economic factors.

[0210] (8) N-BEATS

[0211] N-BEATS is a deep learning model suitable for time series prediction, which is particularly good at capturing long-term dependencies. By using N-BEATS, we can accurately predict the long-term change trend of the recycled amount, and it has good interpretability. N-BEATS can automatically select appropriate features for prediction, avoiding the bias of artificially selecting features.

[0212] 3. Evaluation Metrics of the Algorithm

[0213] The model uses three evaluation metrics to evaluate the model performance: Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Correlation (R2). The definition of each metric is as follows:

[0214] (1) Mean Squared Error (MSE) is a metric to measure the difference between the predicted value and the true value. The smaller its value, the better the performance of the prediction model. The specific formula is as follows:

[0215]

[0216] n is the number of samples, y iis the corresponding true value, is the predicted value.

[0217] (2) Root Mean Square Error (RMSE) is the square root of the mean squared error. It is used, like the mean squared error, to measure the difference between the predicted value and the true value. RMSE has the same unit as the original data and is easier to interpret. The specific formula is as follows:

[0218]

[0219] (3) Correlation R2 represents the degree to which the model explains the data, and its value ranges from 0 to 1. The closer the value is to 1, the stronger the model's ability to explain the data.

[0220]

[0221] 4. Evaluate the effects and usage scenarios of each model

[0222] The dataset of some product recycling stations in the Hong Kong region selected in the present invention, and the specific experimental results Figure 6 are as follows:

[0223] Among them, the selected ones for each region are partial model diagrams.

[0224] Based on the evaluation results of each model on different recycling station datasets, we draw the following conclusions: SVM performs excellently when the change in recycling volume is relatively stable and the dataset is small. The linear assumption of SVM makes it perform well when the data volume is small and the relationship is simple. When the data volume is small and the change in recycling volume is relatively stable, SVM can predict efficiently and accurately. RF can effectively handle non-linear relationships by integrating multiple decision trees and has strong robustness to data noise. It is suitable for scenarios with a moderate amount of data, certain complexity, and noise. Especially when dealing with a large number of features, RF can provide relatively accurate predictions. The decision tree model can make predictions through simple rules and has strong interpretability. In the case of a small amount of data, DT can better capture the relationship between recycling volume and economic factors, especially suitable for preliminary analysis. The BP neural network has strong fitting ability and is suitable for scenarios with strong non-linear relationships, especially when the recycling volume data fluctuates greatly and has complex non-linear change rules. LSTM performs very well in processing time series data and is suitable for datasets with strong time series dependencies. Especially when the change in recycling volume is greatly affected by time factors, LSTM can provide more accurate predictions. TCN performs well in processing time series data with long-term dependencies, has strong parallel processing ability, and can effectively handle long-term dependencies in recycling volume data. It is suitable for large-scale datasets with strong time series characteristics of recycling volume. Especially when the data volume is very large and complex, TCN can effectively improve the calculation efficiency and prediction accuracy. Transformer has a powerful self-attention mechanism and can capture global dependencies. Transformer can effectively improve the prediction accuracy of recycling volume, especially having significant advantages when dealing with large datasets. N-BEATS is a newly emerging time series prediction model in recent years, especially suitable for long time series data. N-BEATS can automatically select appropriate features, avoiding the bias of manual feature selection, and performs excellently on time series data with long-term dependencies. It is suitable for long time series prediction tasks. Especially when the data volume is large and the change trend is complex, N-BEATS can provide stable and high-precision predictions.

[0225] 5. Python-based Tkinter Visualization Interface Settings

[0226] By clicking Figure 7 the recycling volume prediction in Figure 8, on the far left is the artificial intelligence model required for recycling volume prediction, including a single prediction model and a hybrid prediction model; the function below, through the selection area and file import, will import and display the historical recycling volume data of the corresponding area; then above is a visual display of the imported data, including recycling volume, total retail sales of consumer goods, number of tourists, gross domestic product, and per capita disposable income. After importing the data, click the model button to train the imported data using the corresponding model, such as Figure 8 shown as follows:

[0227] After clicking the button, train using the corresponding model and display the training results in a sub-window; the sub-window contains a comparison chart of training values and predicted values and some evaluation indicators including MSE, RMSE, R2, and MAE.

[0228] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated designed hardware. Those of ordinary skill in the art can understand that the above devices and methods can be implemented using computer-executable instructions and / or included in processor control code, such as provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuits of programmable hardware devices such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, or can be implemented by a combination of the above hardware circuits and software such as firmware.

[0229] The above is only the specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention by those skilled in the art within the technical scope disclosed by the present invention shall be covered by the protection scope of the present invention.

Claims

1. A method for predicting the recycling of remanufactured products based on an artificial intelligence model, characterized in that: The following steps are involved: 1) Prediction modeling of remanufactured product recycling: Define the key factors of remanufactured product recycling, including daily gross production value, daily retail sales, etc., and build a recycling volume prediction model to analyze the impact of these factors on remanufactured product recycling and form a basis for predicting waste recycling status; 2) Design and application of multi-algorithm models: Based on the constructed recycling volume prediction model, a variety of artificial intelligence algorithms, including support vector machine (SVM), random forest (RF), decision tree (DT), BP neural network (BP), long short-term memory network (LSTM), temporal convolutional network (TCN), Transformer and N-BEATS model, are applied to predict the recycling volume of remanufactured products; 3) Algorithm performance evaluation: Design evaluation indicators, including mean square error (MSE), mean absolute error (MAE), and R 2 Etc., used to quantify the model's ability to fit the data and the accuracy of the prediction target; 4) Model comparison analysis: Based on the historical data sets of different recycling stations, the prediction results of the above-mentioned multiple artificial intelligence algorithms are compared and analyzed to evaluate the advantages and disadvantages and applicable scope of each algorithm, and provide support for selecting the optimal model for specific scenarios; 5) Visualization of algorithm prediction results: Use Matplotlib to visualize the predicted values ​​of each model and the actual values, intuitively display the deviation between the predicted results and the true values, and present the fitting effect and error distribution of the algorithm through charts; 6) User interface design and application: Build a user interface based on Python Tkinter to display the prediction results and evaluation indicators of different algorithms, support users to customize input data, and view the prediction performance and result comparison of the algorithm in real time.

2. The method according to claim 1, characterized in that The model performance evaluation step includes determining the deviation between the predicted value and the true value by analyzing the mean square error (MSE), quantifying the overall error of the model prediction by the mean absolute error (MAE), and 2 The indicators assess the overall goodness of fit of the model; The design of the multi-algorithm model selects specific models based on different data features, among which random forest is suitable for high-dimensional data feature analysis, long short-term memory network is suitable for time series data prediction, and Transformer model is suitable for long-term prediction tasks.

3. The method according to claim 1, characterized in that The user interface includes an algorithm selection module, a data input module and a result display module, which supports users to compare the prediction effects of multiple algorithms through the interface and adjust the input data in real time.

4. The method according to claim 1, characterized in that: The visualization process of the prediction results includes displaying the comparison between the predicted value and the true value through a line graph, and analyzing the deviation characteristics of the model through a bar graph or an error distribution graph.

5. The method for predicting the recycling of remanufactured products based on an artificial intelligence model as claimed in claim 1, characterized in that: The remanufactured product recycling prediction modeling: First, for the prediction of remanufactured product recycling, the key factors for recycling volume prediction are defined, including economic factors such as daily gross domestic product and daily retail sales. These factors are considered to have a significant impact on the recycling volume of remanufactured products, so they are used as input features of the model. Specifically, the impact of these factors on the recycling volume of remanufactured products is explored, using the following evaluation indicators: The Pearson correlation coefficient was used to explore the correlation between this factor and the recycling of remanufactured products. In addition, the p-value and t-value were combined to evaluate the correlation. The definition of each indicator is as follows: (1) The Pearson correlation coefficient is used to measure the linear correlation between two variables, with a value range of -1 to 1. The specific formula is: Where r is the Pearson correlation coefficient, xi and yi are the observed values ​​of variables X and Y, and is the mean of X and Y; (2) The t value is used to test whether the sample data deviates significantly from the null hypothesis and is usually used for coefficient significance testing in regression analysis. The specific formula is: Where t is the t value, r is the Pearson correlation coefficient, and n is the total number of samples; (3) The p-value is used to evaluate the probability of the observed result occurring when the null hypothesis is true and is an indicator for testing the significance of correlation.

6. The method for predicting the recycling of remanufactured products based on an artificial intelligence model as claimed in claim 2, characterized in that: The design is based on an artificial intelligence model for recycling volume prediction: Use a variety of machine learning algorithms, including support vector machine SVM, random forest RF, decision tree DT, BP neural network BP, long short-term memory network LSTM, temporal convolutional network TCN, Transformer and N-BEATS; (1) Support Vector Machine (SVM) SVM is a typical classification and regression algorithm, which is suitable for predicting small sample and nonlinear data. In this experiment, SVM is used to predict the recovery amount, and different types of input data are processed by selecting appropriate kernel functions such as linear kernel and radial basis kernel. The parameters of SVM are optimized by cross validation and grid search to obtain the best model. (2) Random Forest RF Random forest is an ensemble learning method that improves the accuracy and stability of predictions by integrating multiple decision trees. In the prediction of solid waste recycling, RF can effectively process high-dimensional data and has strong anti-overfitting capabilities. By training multiple decision trees, random forest can capture the complex nonlinear relationship between recycling volume and economic indicators. (3) Decision Tree DT Decision tree is a common supervised learning algorithm that makes predictions by recursively dividing data into different subsets. In the recycling volume prediction task, the decision tree model can better capture the relationship between specific economic factors and recycling volume. (4) BP neural network BP neural network is a common feedforward neural network, which is trained by error back propagation algorithm; In the prediction of solid waste recycling volume, BP network can effectively fit complex nonlinear relationships and perform feature abstraction through hierarchical structure; Optimize the model's predictive ability by continuously adjusting the number of network layers and nodes; (5) Long Short-Term Memory Network (LSTM) LSTM is a recurrent neural network (RNN) that is particularly suitable for processing time series data. The change in the amount of solid waste recycled is time-series, so LSTM can effectively capture the law of changes in the amount of recycled waste over time through its memory units. By training the LSTM network, it is possible to predict future changes in the amount of recycled waste over a long period of time, which is particularly suitable for capturing the lag effect of economic factors. (6) Temporal Convolutional Network (TCN) TCN is a time series data processing method based on convolutional neural network (CNN); unlike LSTM, TCN captures the time series pattern of data through one-dimensional convolution operation, and has higher parallel processing capabilities and longer receptive field.

7. The method for predicting the recycling of remanufactured products based on an artificial intelligence model as claimed in claim 3, characterized in that: The Transformer: Transformer is a deep learning model based on the self-attention mechanism, which is widely used in natural language processing tasks. In recycling volume prediction, Transformer can effectively model the global dependencies between data, especially on large-scale data sets. Through the multi-head self-attention mechanism, Transformer can capture the complex relationship between recycling volume and various economic factors.

8. The method for predicting the recycling of remanufactured products based on an artificial intelligence model as claimed in claim 1, characterized in that: Evaluation index design of the algorithm: The model uses three evaluation indicators to evaluate the model effect: mean square error (MSE), root mean square error (RMSE), and correlation (R2), where each indicator is defined as follows: (1) Mean square error (MSE) is an indicator that measures the difference between the predicted value and the true value. The smaller the value, the better the performance of the prediction model. The specific formula is as follows: n is the number of samples, y i is the corresponding true value, is the predicted value; (2) Root mean square error (RMSE) is the square root of the mean square error. Like the mean square error, it is used to measure the difference between the predicted value and the true value. RMSE has the same unit as the original data and is easier to interpret. The specific formula is as follows: (3) Correlation R2 indicates the degree to which the model explains the data, and its value range is between 0 and 1; the closer the value is to 1, the stronger the model's ability to explain the data; 9. A system for predicting the recycling of remanufactured products based on an artificial intelligence model, which implements the method for predicting the recycling of remanufactured products based on an artificial intelligence model as claimed in any one of claims 1 to 8, characterized in that: The remanufactured product recycling prediction system based on the artificial intelligence model includes: Predictive modeling module, used for predictive modeling of remanufactured product recycling: first define the key factors of remanufactured product recycling, and generate waste recycling status by establishing a recycling volume prediction model; build a remanufactured product recycling prediction model that includes daily gross domestic product and daily retail sales factors, the main purpose of which is to study the impact of these factors on remanufactured product recycling; algorithm, in this experiment alone, design an artificial intelligence model based on recycling volume prediction, using a variety of machine learning algorithms, including support vector machine SVM, random forest RF, decision tree DT, BP neural network BP, long short-term memory network LSTM, time convolution network TCN, Transformer, N-BEATS; Index design module, used for algorithm evaluation index design: design evaluation indicators for the algorithm to intuitively reflect the degree of fit of the algorithm to the data and the accuracy of the prediction of the target motion elements; The comparative analysis module is used to compare and analyze the performance of these models on different recycling bin data sets to evaluate the advantages, disadvantages and applicability of each model; Visualization module, used for visualization of algorithm predictions: Use matplotlib to visualize algorithm predictions, and use charts to clearly show the comparison between the prediction results of each algorithm and the actual value; Interface design module, used to create user interfaces based on Tkinter in Python for visual display of remanufacturing product recycling; providing instant algorithm comparison and data display.

10. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the remanufactured product recycling prediction system based on the artificial intelligence model as described in claim 7.

Images