A waste product residual value evaluation method based on double-channel pricing

Abstract

The application discloses a kind of based on double channel pricing's waste and old article residual value evaluation method, it is related to article recycling technical field, including the following steps: pre-acquisition user input attribute data and image data are extracted respectively, generate structured attribute feature vector and image appearance feature vector, and splice into standardization comprehensive feature vector;Fuzziness, exposure, reflection ratio, main body ratio are calculated to each image.The application realizes the intelligent accurate shunting of waste channel and second-hand channel by introducing the dynamic classification mechanism of residual value quantitative determination and periodically updating category threshold, effectively solves the cross-border article mismatching problem, significantly reduces the second negotiation rate of going to the door, improves the recovery efficiency and user trust degree, and realizes the real-time self-adaptive adjustment of waste price to upstream market fluctuation, regional supply and demand difference and time period transport capacity change, significantly improves the accuracy, timeliness and scene adaptability of waste pricing.

Description

Technical Field

[0001] This invention relates to the field of recycling technology, and more specifically, to a method for assessing the residual value of waste materials based on dual-channel pricing. Background Technology

[0002] As the circular economy system continues to improve, the waste recycling industry is gradually transforming from offline, fragmented collection to internet-based, platform-based, and intelligent transformation. The market is now characterized by a hybrid business model where waste recycling and secondhand goods circulation coexist. Existing recycling platforms mostly employ a separate pricing structure, with the waste and secondhand ends operating independently, lacking data sharing and mechanisms, and failing to form a unified, intelligent pricing system, making it difficult to adapt to the actual needs of the current hybrid recycling scenario.

[0003] Current pricing technologies in mixed recycling scenarios generally suffer from five core flaws: rigid pricing, reliance on subjectivity, chaotic sorting, weak interpretability, and difficulty in iteration. First, waste recycling pricing often employs fixed unit prices or infrequent manual price adjustments, failing to respond in real-time to dynamic factors such as fluctuations in upstream raw material markets, regional supply and demand differences, and changes in transportation capacity during specific times, resulting in poor timeliness and significant deviations in pricing. Second, secondhand item valuation heavily relies on manual experience or simple rule-based tables, lacking quantitative modeling of individual differences in item appearance, function, and accessory integrity, leading to high dispersion and poor consistency in valuation results for items of the same model. Third, for cross-border items bordering on repairability and non-repairability, existing technologies lack a unified, quantitative, and automated sorting and judgment mechanism, easily leading to mismatches between waste and secondhand channels, resulting in frequent secondary price negotiations after door-to-door collection and decreased user trust. Fourth, pricing results lack transparent breakdown criteria, making it difficult for users to understand the price composition logic, resulting in insufficient platform pricing interpretability and low conversion rates.

[0004] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0005] In response to the problems in related technologies, this invention proposes a method for assessing the residual value of waste materials based on dual-channel pricing, in order to overcome the aforementioned technical problems existing in the existing related technologies.

[0006] The technical solution of this invention is implemented as follows:

[0007] A method for assessing the residual value of waste materials based on dual-channel pricing includes the following steps:

[0008] The attribute data and image data input by the user are collected in advance, and feature extraction is performed separately to generate structured attribute feature vectors and image quality feature vectors, which are then concatenated into a standardized comprehensive feature vector.

[0009] For each image, the blur, exposure, reflection ratio, and subject ratio are calculated. Qualified images are selected using a quality judgment function. If the number of qualified images is less than the minimum threshold, a reshoot is triggered.

[0010] Make dynamic classification decisions, calculate the remaining value of items, set dynamic classification thresholds according to the categories of items, and compare the remaining value of items with the classification thresholds to determine whether to enter the waste channel or the second-hand channel;

[0011] The system outputs the pricing results for the corresponding channel based on the classification results, breaks down the price components, and prompts the user to retake the photo or triggers manual review when the confidence level is low.

[0012] Furthermore, the attribute feature vector The image quality feature vector is generated by encoding and normalizing the product category, brand, model, usage status, geographical location, and user profile. The image is generated by extracting and pooling valid images from multiple angles uploaded by users.

[0013] Furthermore, the quality determination function , is represented as:

[0014] ;

[0015] In the formula, The quality indicator for the i-th image is deemed acceptable; I(⋅) is the indicator function; The threshold for acceptable ambiguity; , These are the lower and upper limits of the acceptable exposure range; The acceptable threshold for reflectivity ratio; The threshold for the proportion of the main body is qualified.

[0016] Furthermore, the calculation of the residual value of the item is expressed as:

[0017] ;

[0018] In the formula, The baseline residual value for model m. This is a comprehensive correction factor for appearance and attributes. This is the image quality feature vector. This is a structured attribute feature vector.

[0019] Furthermore, comparing the remaining value of an item with a classification threshold includes:

[0020] like Then it enters the waste disposal channel;

[0021] like Then it will enter the second-hand channel;

[0022] in, This is the classification threshold.

[0023] Furthermore, the classification threshold This includes: periodic updates, represented as:

[0024] ;

[0025] In the formula, Let be the classification threshold for category c in period t; The updated classification threshold for category c in period t+1; The learning rate for error routing bias and bargaining rate bias; This represents the recent misclassification deviation value for product category c. This represents the recent price negotiation rate deviation for product category c.

[0026] Furthermore, the process of entering the waste channel includes calculating the waste assessment price, expressed as:

[0027] ;

[0028] In the formula, Let k be the market benchmark unit price for waste subcategories. For the supply and demand adjustment factor of region g at time t, Let W be the value coefficient for user u, and W be the weight of the item.

[0029] Among them, supply and demand adjustment factors , represented as:

[0030] ;

[0031] In the formula, α is the supply and demand adjustment intensity coefficient; tanh(⋅) is the hyperbolic tangent function; For region g and time period t, the number of recycling orders; For region g and time period t, the number of online recycling workers; It is a very small constant; This serves as the benchmark threshold for supply and demand balance.

[0032] Among them, the user value coefficient , is represented as:

[0033] ;

[0034] In the formula, β represents the user coefficient intensity; S u A comprehensive historical rating for user u; The average rating of all users on the platform.

[0035] Furthermore, the entry into the secondary market includes constructing a training set of historical transactions. Using a regression model , is represented as:

[0036] ;

[0037] In the formula, For basic valuation of second-hand items; f(⋅) is the trained regression prediction model; X is the comprehensive feature vector of the item to be evaluated; X i Let y be the feature vector of the i-th historical sample; i Let N be the transaction price of the i-th historical sample; N is the total number of training samples.

[0038] The objective function of the model is expressed as:

[0039] ;

[0040] In the formula, Let $i$ be the predicted price of the model for the $i$-th sample. This is a regularization term for the model to prevent overfitting.

[0041] For cold-start models, a reference price is obtained through the fusion of similar samples. This leads to the basic valuation after fusion. , is represented as:

[0042] ;

[0043] Among them, reference price , is represented as:

[0044] ;

[0045] In the formula, Reference prices for similar samples; The set of historical samples with similar characteristics to the item to be evaluated; exp(⋅) is an exponential function; The base valuation for secondhand properties after fusion; λ is the model prediction weight coefficient; Let the weight of the j-th similar sample be denoted as: ,That Let be the feature distance between the item to be evaluated and the j-th sample.

[0046] Furthermore, this also includes: conducting a condition assessment and price adjustment, including the following steps:

[0047] Extract multi-branch damage scores from valid images, including: screen damage score D. s ; Casing wear score D b Interface button damage score D pLens damage score D c Attachment completeness score: D a ;

[0048] The overall condition score is calculated and expressed as follows:

[0049] ;

[0050] In the formula, These are the weighting coefficients for the screen, casing, interface, lens, and accessories, respectively.

[0051] The overall condition score is mapped to a condition correction factor to calculate the final used price and price range, expressed as follows:

[0052] ;

[0053] ;

[0054] In the formula, The final valuation price for the secondary market; To assess the lower and upper limits of the price range; δ is the range expansion coefficient, with values ​​ranging from... ;

[0055] Among them, the condition correction factor , is represented as:

[0056] ;

[0057] In the formula, C cond This is a condition correction factor; Here, γ is the Sigmoid activation function; γ is the correction coefficient mapping slope. This serves as the baseline quality threshold. To correct the lower and upper limits of the coefficient.

[0058] The beneficial effects of this invention are:

[0059] This invention achieves intelligent and precise diversion between waste and secondhand channels by introducing a dynamic classification mechanism for quantifying residual value and periodically updating category thresholds. This effectively solves the problem of mismatched cross-border goods, significantly reduces the rate of secondary price negotiations at the door, and improves recycling efficiency and user trust. Through a multi-factor linkage pricing model based on the waste market benchmark price, supply and demand adjustment factors, user value coefficient, and weight, the invention enables real-time adaptive adjustment of waste prices to upstream market fluctuations, regional supply and demand differences, and changes in transportation capacity during specific time periods. This significantly improves the accuracy, timeliness, and scenario adaptability of waste pricing. Finally, by training a regression model using standardized comprehensive feature vectors from the secondhand side to output a basic valuation and combining this with computer vision to extract multi-dimensional condition damage features and generate correction coefficients, the invention achieves refined valuation of the residual value of secondhand goods. The system offers several key advantages: Quantifiable assessment, eliminating reliance on manual experience and improving the consistency and accuracy of pricing; reliable pricing for new categories and cold-start models achieved through similar sample fusion strategies, expanding the applicability and scenario coverage of the method; transparent breakdown of price composition logic through interpretable pricing result display design, enhancing pricing transparency and user acceptance, and effectively improving conversion rates; furthermore, a closed-loop feedback and iteration mechanism for data collection, indicator evaluation, periodic retraining, incremental updates, and A / B testing with gray-scale releases enables continuous dynamic optimization of model parameters, classification thresholds, and pricing strategies, ensuring the long-term stability, timeliness, and generalization ability of the pricing model, reducing manual operating costs, and promoting the efficient development of the waste recycling industry towards large-scale, standardized, and intelligent operations. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a flowchart illustrating a method for assessing the residual value of waste materials based on dual-channel pricing, according to an embodiment of the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0063] According to an embodiment of the present invention, a method for assessing the residual value of waste materials based on dual-channel pricing is provided.

[0064] like Figure 1 As shown, the method for assessing the residual value of waste materials based on dual-channel pricing according to an embodiment of the present invention includes the following steps:

[0065] Step S1: Pre-collect user-input attribute data and image data, extract features respectively, and generate structured attribute feature vectors. Image quality feature vector And concatenate them into a standardized comprehensive feature vector. ;

[0066] Among them, attribute feature vector Image quality feature vector is generated by encoding and normalizing structured information such as product category, brand, model, usage status, geographical location, and user profile. The image is generated by extracting and pooling valid images from multiple angles uploaded by users.

[0067] Specifically, there should be multiple effective images, i.e., no fewer than 3, preferably 5, multi-angle photos of the object;

[0068] Step S2: Calculate the blur level B for each image. i Exposure E i Reflectivity R i , Main body proportion O i Through the quality judgment function If the number of qualified images is less than the minimum threshold N, then the process is repeated. min If so, a replay will be triggered; otherwise, proceed to the next step.

[0069] Among them, the quality judgment function , is represented as:

[0070] ;

[0071] In the formula, This is the quality indicator for the i-th image; I(⋅) is an indicator function, which takes the value 1 if the condition in parentheses is true, and 0 otherwise; The threshold for acceptable ambiguity; , These are the lower and upper limits of the acceptable exposure range; The acceptable threshold for reflectivity ratio; The threshold for the proportion of the main body to meet the qualification requirements;

[0072] Specifically, ambiguity B i This is the normalized value for image edge energy, ranging from [0,1]. A larger value indicates a more blurred image. Exposure E iThis is the normalized grayscale mean value of the image, ranging from [0,1], used to determine whether the image is too dark, normal, or overexposed. Reflectance R i This represents the proportion of pixels in the high-brightness reflective area to the total number of pixels in the image, with a value ranging from [0,1]. A larger value indicates more severe reflection. Subject proportion O i This represents the proportion of pixels in the main area of ​​the object to the total number of pixels in the image. The value ranges from [0,1]. A larger value indicates that the object occupies a higher proportion and is more complete in the image.

[0073] Step S3 involves making a dynamic classification decision, including the following steps:

[0074] Step S201, calculate the remaining value of the item, expressed as:

[0075] ;

[0076] In the formula, The baseline residual value for model m. This is a comprehensive correction factor for appearance and attributes. This is the image quality feature vector. For structured attribute feature vectors;

[0077] Step S202: Set a dynamic classification threshold according to the category c of the item. ;

[0078] Among them, if Then it enters the waste disposal channel;

[0079] like Then it will enter the second-hand channel;

[0080] Specifically, during implementation, the classification threshold... Periodic updates are represented as:

[0081] ;

[0082] In the formula, Let be the classification threshold for category c in period t; The updated classification threshold for category c in period t+1; The learning rate for error routing bias and bargaining rate bias; This represents the recent misclassification deviation value for product category c. This represents the recent price negotiation rate deviation for product category c.

[0083] Step S3: Set a waste channel price to calculate the waste assessment price, expressed as:

[0084] ;

[0085] In the formula, Let k be the market benchmark unit price for waste subcategories. For the supply and demand adjustment factor of region g at time t, Let W be the value coefficient for user u, and W be the weight of the item.

[0086] Specifically, supply and demand adjustment factors , is represented as:

[0087] ;

[0088] In the formula, α is the supply and demand adjustment intensity coefficient. ; tanh(⋅) is the hyperbolic tangent function, used to normalize the supply and demand imbalance; For region g and time period t, the number of recycling orders; For region g and time period t, the number of online recycling workers; It is a very small constant to prevent the denominator from being 0; This serves as the benchmark threshold for supply and demand balance.

[0089] User value coefficient , is represented as:

[0090] ;

[0091] In the formula, β represents the user coefficient intensity. S u A comprehensive historical rating for user u; The average rating of all users on the platform.

[0092] Step S4 involves pricing the secondary market channel for valuation, including constructing a training set of historical transactions. GBDT / XGBoost / LightGBM regression models were used. , is represented as:

[0093] ;

[0094] In the formula, For basic valuation of second-hand items; f(⋅) is the trained regression prediction model; X is the comprehensive feature vector of the item to be evaluated; X i Let y be the feature vector of the i-th historical sample; i Let N be the transaction price of the i-th historical sample; N is the total number of training samples.

[0095] The objective function of the model is expressed as:

[0096] ;

[0097] In the formula, Let $i$ be the predicted price of the model for the $i$-th sample. This is a regularization term for the model to prevent overfitting.

[0098] Specifically, for cold-start models, a reference price is obtained by fusing similar samples. This leads to the basic valuation after fusion. , is represented as:

[0099] ;

[0100] Among them, reference price , is represented as:

[0101] ;

[0102] In the formula, Reference prices for similar samples; The set of historical samples with similar characteristics to the item to be evaluated; exp(⋅) is an exponential function; This represents the base valuation of secondhand properties after fusion; λ is the model prediction weight coefficient. ; The weight of the j-th similar sample is expressed as: ,That The feature distance between the item to be evaluated and the j-th sample;

[0103] Step S5 involves assessing the condition and adjusting the price, as detailed below:

[0104] Step S501: Extract multi-branch damage scores from the valid image, including: screen damage score D. s Values ​​range from 0 (no damage) to 1 (severe damage); shell wear score D b Values ​​range from 0 to 1; Interface button damage score D p Values ​​range from 0 to 1; Lens damage score D c Values ​​range from 0 to 1; Attachment completeness score: D a The value 0 indicates no accessories, while 1 indicates all accessories are included.

[0105] Step S502: Calculate the overall condition score, with a value of 0 representing the worst condition and 1 representing brand new. This is represented as:

[0106] ;

[0107] In the formula, These are the weighting coefficients for the screen, casing, interface, lens, and accessories, respectively.

[0108] Step S503: Map the overall condition score to a condition correction factor, calculate the final used price and price range, as follows:

[0109] ;

[0110] ;

[0111] In the formula, The final valuation price for the secondary market; To assess the lower and upper limits of the price range; δ is the range expansion coefficient, with values ​​ranging from... ;

[0112] Among them, the condition correction factor , is represented as:

[0113] ;

[0114] In the formula, C cond This is a condition correction factor; The Sigmoid activation function outputs 0 to 1; γ is the correction coefficient and mapping slope. This serves as the baseline quality threshold. To correct the lower and upper limits of the coefficient.

[0115] Step S6: Output the pricing results for the corresponding channel based on the classification results, break down the price composition, and prompt the user to retake the photo or trigger manual review when the confidence level is low.

[0116] Specifically, the scrap channel displays the market benchmark price, supply and demand factors, user coefficient, weight, and final price; the second-hand channel displays the basic valuation, condition correction coefficient, final price, price range, and confidence level.

[0117] In addition, the following steps are also included:

[0118] Step S7: Collect user behavior, on-site verification price, transaction price, reasons for negotiation, and referral results data, and analyze them using the Mean Absolute Percentage Error (MAPE) and secondary negotiation rate. To evaluate the pricing effectiveness, the base model is periodically retrained. Incrementally update thresholds and parameters, and complete model iteration through A / B testing with canary releases.

[0119] The mean absolute percentage error (MAPE) is expressed as follows:

[0120] ;

[0121] Among them, the second negotiation rate , is represented as:

[0122] ;

[0123] In the formula, Number of negotiated orders; This represents the number of completed orders.

[0124] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Those skilled in the art, upon considering the disclosure in the specification and embodiments, will readily conceive of other embodiments of this disclosure. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

[0125] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for assessing the residual value of waste materials based on dual-channel pricing, characterized in that, Includes the following steps: The attribute data and image data input by the user are collected in advance, and feature extraction is performed separately to generate structured attribute feature vectors and image quality feature vectors, which are then concatenated into a standardized comprehensive feature vector. For each image, the blur, exposure, reflection ratio, and subject ratio are calculated. Qualified images are selected using a quality judgment function. If the number of qualified images is less than the minimum threshold, a reshoot is triggered. Make dynamic classification decisions, calculate the remaining value of items, set dynamic classification thresholds according to the categories of items, and compare the remaining value of items with the classification thresholds to determine whether to enter the waste channel or the second-hand channel; The system outputs the pricing results for the corresponding channel based on the classification results, breaks down the price components, and prompts the user to retake the photo or triggers manual review when the confidence level is low.

2. The method for assessing the residual value of waste materials based on dual-channel pricing according to claim 1, characterized in that, The attribute feature vector The image quality feature vector is generated by encoding and normalizing the product category, brand, model, usage status, geographical location, and user profile. The image is generated by extracting and pooling valid images from multiple angles uploaded by users.

3. The method for assessing the residual value of waste materials based on dual-channel pricing according to claim 1, characterized in that, The quality judgment function , is represented as: ； In the formula, The quality indicator for the i-th image is deemed acceptable; I(⋅) is the indicator function; The threshold for acceptable ambiguity; , These are the lower and upper limits of the acceptable exposure range; The acceptable threshold for reflectivity ratio; The threshold for the proportion of the main body is qualified.

4. The method for assessing the residual value of waste materials based on dual-channel pricing according to claim 1, characterized in that, The calculation of the residual value of the item is expressed as follows: ； In the formula, The baseline residual value for model m. This is a comprehensive correction factor for appearance and attributes. This is the image quality feature vector. This is a structured attribute feature vector.

5. The method for assessing the residual value of waste materials based on dual-channel pricing according to claim 4, characterized in that, The comparison of the remaining value of an item with a classification threshold includes: like Then it enters the waste disposal channel; like Then it will enter the second-hand channel; in, This is the classification threshold.

6. The method for assessing the residual value of waste materials based on dual-channel pricing according to claim 5, characterized in that, The classification threshold This includes: periodic updates, represented as: ； In the formula, Let be the classification threshold for category c in period t; The updated classification threshold for category c in period t+1; The learning rate for error routing bias and bargaining rate bias; This represents the recent misclassification deviation value for product category c. This represents the recent price negotiation rate deviation for product category c.

7. The method for assessing the residual value of waste materials based on dual-channel pricing according to claim 6, characterized in that, The process of entering the waste channel includes calculating the waste assessment price, expressed as: ； In the formula, Let k be the market benchmark unit price for waste subcategories. For the supply and demand adjustment factor of region g at time t, Let W be the value coefficient for user u, and W be the weight of the item. Among them, supply and demand adjustment factors , is represented as: ； In the formula, α is the supply and demand adjustment intensity coefficient; tanh(⋅) is the hyperbolic tangent function; For region g and time period t, the number of recycling orders; For region g and time period t, the number of online recycling workers; It is a very small constant; This serves as the benchmark threshold for supply and demand balance. Among them, the user value coefficient , is represented as: ； In the formula, β represents the user coefficient intensity; S u A comprehensive historical rating for user u; The average rating of all users on the platform.

8. The method for assessing the residual value of waste materials based on dual-channel pricing according to claim 6, characterized in that, The entry into the secondary market includes constructing a training set of historical transactions. Using a regression model , is represented as: ； In the formula, For basic valuation of second-hand items; f(⋅) is the trained regression prediction model; X is the comprehensive feature vector of the item to be evaluated; X i Let y be the feature vector of the i-th historical sample; i Let N be the transaction price of the i-th historical sample; N is the total number of training samples. The objective function of the model is expressed as: ； In the formula, Let $i$ be the predicted price of the model for the $i$-th sample. For model regularization terms; For cold-start models, a reference price is obtained through the fusion of similar samples. This leads to the basic valuation after fusion. , is represented as: ； Among them, reference price , is represented as: ； In the formula, Reference prices for similar samples; The set of historical samples with similar characteristics to the item to be evaluated; exp(⋅) is an exponential function; The base valuation for secondhand properties after fusion; λ is the model prediction weight coefficient; The weight of the j-th similar sample is expressed as: ,That Let be the feature distance between the item to be evaluated and the j-th sample.

9. The method for assessing the residual value of waste materials based on dual-channel pricing according to claim 8, characterized in that, Also includes: Adjusting the price based on condition involves the following steps: Extract multi-branch damage scores from valid images, including: screen damage score D. s ; Casing wear score D b Interface button damage score D p Lens damage score D c Attachment completeness score: D a ; The overall condition score is calculated and expressed as follows: ； In the formula, These are the weighting coefficients for the screen, casing, interface, lens, and accessories, respectively. The overall condition score is mapped to a condition correction factor to calculate the final used price and price range, expressed as follows: ； ； In the formula, The final valuation price for the secondary market; To assess the lower and upper limits of the price range; δ is the range expansion coefficient, with values ​​ranging from... ; Among them, the condition correction factor , represented as: ； In the formula, C cond This is a condition correction factor; Here, γ is the Sigmoid activation function; γ is the correction coefficient mapping slope. This serves as the baseline quality threshold. To correct the lower and upper limits of the coefficient.

Images