Search

CN-121279157-B - Multi-mode residual value evaluation method and system for mobile terminal pre-deduction

CN121279157BCN 121279157 BCN121279157 BCN 121279157BCN-121279157-B

Abstract

The invention discloses a multi-mode residual value evaluation method and a multi-mode residual value evaluation system for mobile terminal pre-deduction, which belong to the field of electric digital processing and comprise the following steps of movable characteristic acquisition, movable characteristic reliability evaluation, residual value evaluation result verification and mobile terminal pre-deduction. According to the invention, static data are input into the migration learning model to be analyzed to obtain the migration characteristics, reliability evaluation is synchronously carried out on the migration characteristics, whether reliability adjustment measures are adopted or not is determined based on the evaluation results, then the residual value evaluation is carried out to output the residual value analysis results to carry out reliability verification, the mobile terminal to be evaluated is subjected to pre-deduction based on the residual value analysis results after the reliability verification, whether the lease is ended or not is determined based on a triggered reservation mechanism of the mobile terminal to be evaluated after the lease is preset, the accuracy of the pre-deduction data is improved, and the problem that the accuracy of the pre-deduction results is reduced due to the fact that the reliability of the residual value evaluation results is reduced in the prior art is solved.

Inventors

  • YU FEI

Assignees

  • 天津奇立软件技术有限公司

Dates

Publication Date
20260508
Application Date
20251211

Claims (9)

  1. 1. The multi-mode residual value evaluation method for the mobile terminal pre-deduction is characterized by comprising the following steps of: inputting static data covering the functional attribute of the mobile terminal to be evaluated into a transfer learning model, and analyzing the corresponding transferable characteristics representing the damage degree of the mobile terminal to be evaluated after the preset lease; performing reliability evaluation on the movable characteristics synchronously, and determining whether to take corresponding reliability adjustment measures based on corresponding evaluation results, wherein the reliability adjustment measures comprise image characteristic deficiency compensation measures, hardware performance characteristic correction measures and abrasion characteristic correction measures; Performing residual value evaluation according to the movable characteristics and static data of the mobile terminal to be evaluated to output a corresponding residual value analysis result, and performing reliability verification on the residual value analysis result; The specific process for verifying the reliability of the residual value analysis result comprises the following steps: inputting the movable characteristics and static data of the mobile terminal to be evaluated into a transfer learning model to output corresponding residual analysis data, wherein the residual analysis data comprises residual prediction results, depreciation curves, available deduction amounts and confidence coefficients; if the confidence coefficient is higher than the preset confidence coefficient, carrying out residual value evaluation cross verification, otherwise, carrying out pre-deduction based on the currently output residual value analysis data; The specific contents of the residual value evaluation cross-validation are as follows: Inputting the movable characteristics and static data of the mobile terminal to be evaluated into a standby RFR residual value verification model and a standby LM-BP residual value verification model, respectively outputting corresponding residual value analysis data, and respectively recording the residual value analysis data as RFR residual value analysis data and LM-BP residual value analysis data; if the confidence coefficient of the RFR residual value analysis data and the LM-BP residual value analysis data is higher than the preset confidence coefficient, respectively mapping the confidence coefficient of the RFR residual value analysis data and the LM-BP residual value analysis data to obtain corresponding confidence coefficient weights based on the confidence coefficient of the RFR residual value analysis data and the LM-BP residual value analysis data, coupling the RFR residual value analysis data and the LM-BP residual value analysis data based on the confidence coefficient weights, and performing pre-deduction according to the residual value analysis data obtained by coupling; If the confidence coefficient in the RFR residual value analysis data and the LM-BP residual value analysis data is not higher than the preset confidence coefficient at the same time, performing pre-deduction by adopting residual value analysis data with the confidence coefficient higher than the corresponding preset confidence coefficient; If the confidence coefficient of the RFR residual value analysis data and the LM-BP residual value analysis data is not higher than the preset confidence coefficient, respectively obtaining the difference of the confidence coefficient of the residual value analysis data, the RFR residual value analysis data and the LM-BP residual value analysis data, mapping the difference to obtain a corresponding migration learning sample increment, sequentially increasing the migration learning sample increment according to the migration learning sample increment to retrain a migration learning model, synchronously continuously performing residual value assessment based on the migration learning characteristics output after the migration learning sample increment until the confidence coefficient of the output residual value analysis data is higher than the preset confidence coefficient, otherwise sequencing all the obtained residual value analysis data according to the confidence coefficient, and pre-deducting the residual value analysis data with the highest confidence coefficient; and pre-deducting the mobile terminal to be evaluated based on the residual value analysis result after the credibility verification, and determining whether to end the lease based on a triggered reservation mechanism of the mobile terminal to be evaluated after the lease is preset.
  2. 2. The method for evaluating the multi-modal residual values for mobile terminal pre-deduction of claim 1, wherein the migratable features include image features, hardware performance features and wear features; the image features comprise the number of scratches, the length of the scratches, the depth of the scratches, the area occupation ratio of the scratches, the broken screen grade, the detection result of the bright spots and the broken spots of the screen, the color cast degree of the screen and the number of scratches of the camera; the hardware performance characteristics comprise CPU performance attenuation, GPU performance attenuation, system blocking rate, background application quantity, performance mode starting frequency and cold start time consumption; the wear characteristics include battery health, measured maximum capacity, charge cycle number, average charge temperature, speaker volume decay, fingerprint identification success rate, and face recognition failure rate.
  3. 3. The method for evaluating the multi-modal residual values for the mobile terminal pre-deduction of claim 1, wherein the reliability evaluation comprises the steps of performing category distribution analysis, hardware performance characteristic qualification analysis and wear characteristic correlation analysis on image characteristics; The specific flow of the category distribution analysis of the image features is as follows: reading image features of the mobile terminal to be evaluated, which are output by the transfer learning model, and counting to obtain the number of the surface defect types of the mobile terminal characterized by each image feature, if the number meets the preset defect type labeling number range, the image feature type diversity is reliable, and the image feature distribution uniformity is judged; if the number does not meet the preset defect type labeling number range, adopting image feature deficiency compensation measures, specifically, comparing the obtained defect types and number of the surface of the missing mobile terminal, respectively marking the defect types and the defect compensation image feature number, mapping the obtained image sample learning increment based on the defect compensation image feature number, and retraining the migration learning model by increasing the image feature sample number corresponding to the defect compensation image feature type.
  4. 4. The method for evaluating the multimode residual value for the mobile terminal pre-deduction according to claim 3, wherein the specific process for judging the uniformity of the image characteristic distribution is as follows: Performing duty ratio analysis on the types and the quantities of the surface defects of the mobile terminal to obtain the characteristic duty ratio of each surface defect image; And carrying out difference operation on the duty ratio of each surface defect image characteristic to obtain a corresponding duty ratio difference value, if the duty ratio difference value is larger than the set maximum difference value, recording the difference operation result between the duty ratio difference value and the maximum difference value as a weight adjustment degree value, and mapping based on the weight adjustment degree value to obtain the sample size weight of the corresponding surface defect image characteristic, thereby increasing the sample number of the corresponding surface defect image characteristic so as to retrain the migration learning model.
  5. 5. The method for evaluating the multimode residual value for the mobile terminal pre-deduction according to claim 3, wherein the specific process of the hardware performance characteristic qualification analysis is as follows: Comparing the hardware performance characteristics with a preset industry standard variation range, if the hardware performance characteristics do not belong to the corresponding industry standard variation range, adopting hardware performance characteristic correction measures, carrying out mean value operation on the deviation values of the hardware performance characteristics and the corresponding industry standard variation range to obtain total hardware performance deviation values, and mapping to obtain hardware performance data sample increment amplitude values according to the total hardware performance deviation values, so that the sample quantity corresponding to the hardware performance characteristics is increased to retrain the migration learning model; If the hardware performance characteristics do not belong to the industry standard variation range after the hardware performance data sample increment amplitude value is optimized, increasing the sample quantity corresponding to the hardware performance characteristics based on the preset times until the hardware performance characteristics all belong to the industry standard variation range, otherwise, weighting and coupling are carried out based on the total hardware performance deviation value of the preset times to obtain a hardware performance correction quantity, and the corresponding hardware performance characteristics are corrected based on the correction quantity.
  6. 6. The method for evaluating the multimode residual values for the mobile terminal pre-deduction according to claim 3, wherein the specific steps of the abrasion characteristic correlation analysis are as follows: Based on the output wear characteristics, obtaining a wear performance reliability correlation coefficient of the mobile terminal to be evaluated after a preset lease, and judging with a set reliability threshold value: if the wear performance reliability correlation coefficient is higher than the reliability threshold, the wear characteristic data is reliable, and no additional processing is performed; And if the wear performance reliability correlation coefficient is not higher than the reliability threshold, the wear characteristic data is unreliable, and a wear characteristic correction measure is adopted to correct the wear characteristic.
  7. 7. The method for evaluating the multimode residual value for the mobile terminal pre-deduction according to claim 6, wherein the specific process of adopting the abrasion characteristic correction measures to correct the abrasion characteristics is as follows: The method comprises the steps of simulating the wear characteristic performance of a mobile terminal to be evaluated, and outputting corresponding simulated wear characteristics; carrying out data extraction processing on the simulated wear characteristics according to a preset repair proportion to obtain corresponding repair simulation characteristics; carrying out data elimination processing on the abrasion characteristics based on a preset repair proportion to obtain corresponding abrasion characteristics to be corrected; and carrying out compensation operation on the repair simulation characteristic and the abrasion characteristic to be corrected to obtain a corrected abrasion characteristic, and replacing the abrasion characteristic through the corrected abrasion characteristic.
  8. 8. The method for evaluating the multimode residual value for the mobile terminal pre-deduction of claim 1, wherein the determination of whether to end the lease is based on a reservation mechanism triggered by the mobile terminal to be evaluated is as follows: If the triggered reservation mechanism is due and returns, a recovery process is initiated, wherein the recovery process means that a recovery manufacturer carries out detection and evaluation on the mobile terminal to be evaluated according to a preset standard, if the mobile terminal to be evaluated accords with the contracted recovery standard, the completion of lease performance is judged, otherwise, a claim payment request is sent; If the triggered reservation mechanism is used for reserving the mobile terminal to be evaluated, a pre-deduction amount payment request is initiated, and if the payment request is completed, the lease intelligent contract is terminated and ownership of the mobile terminal to be evaluated is transferred.
  9. 9. A system for applying the mobile terminal pre-deduction-oriented multi-mode residual value assessment method according to any one of claims 1 to 8, which is characterized by comprising a migratable feature acquisition module, a migratable feature reliability assessment module, a residual value assessment result verification module and a mobile terminal pre-deduction module; The mobile terminal comprises a mobile terminal to be evaluated, a movable characteristic acquisition module, a movable characteristic analysis module and a movable characteristic analysis module, wherein the movable characteristic acquisition module is used for inputting static data covering all aspects of functional attributes of the mobile terminal to be evaluated into a movable learning model to analyze corresponding movable characteristics representing the damage degree of the mobile terminal to be evaluated after a preset lease; The movable characteristic reliability evaluation module is used for synchronously carrying out reliability evaluation on the movable characteristics and determining whether to take corresponding reliability adjustment measures based on corresponding evaluation results, wherein the reliability adjustment measures comprise image characteristic deficiency compensation measures, hardware performance characteristic correction measures and abrasion characteristic correction measures so as to ensure the reliability degree of sample input of the movable learning model; the residual value evaluation result verification module is used for carrying out residual value evaluation according to the movable characteristics and static data of the mobile terminal to be evaluated, outputting a corresponding residual value analysis result, and carrying out credibility verification on the residual value analysis result; The mobile terminal pre-deduction module is used for pre-deducting the mobile terminal to be evaluated based on the residual value analysis result after the credibility verification, and determining whether to end the lease based on a triggered reservation mechanism of the mobile terminal to be evaluated after the lease is preset.

Description

Multi-mode residual value evaluation method and system for mobile terminal pre-deduction Technical Field The invention relates to the technical field of electric digital processing, in particular to a mobile terminal pre-deduction-oriented multi-mode residual value evaluation method and system. Background In order to standardize the mobile phone value, reduce the manual detection cost to support automatic recovery and improve the old and new efficiency, a mobile phone residual value evaluation system is needed, which generally adopts a technical framework of multi-mode fusion, large-scale historical data learning and rule calibration, and comprises scoring the hardware configuration and appearance damage of the mobile phone through preset rules, performing residual value evaluation based on a residual value prediction model driven by machine learning or data, and performing intelligent automatic evaluation through multi-mode deep learning. The mobile phone residual value assessment method based on the blockchain is disclosed in China patent with a publication number of CN116894278A and comprises the steps of periodically and autonomously collecting component information of all components in a mobile phone and performing hash calculation on the component information to generate hash values of all the components, storing the hash values of all the components generated at present in association with an identification code uniquely identifying the mobile phone into a blockchain by the blockchain access client after generating the hash values of all the components at each time, responding to information query operation of the mobile phone, inquiring the blockchain by the blockchain access client, acquiring hash values of all the components stored in association with the identification code of the mobile phone from the blockchain, and determining and outputting the mobile phone residual value according to the change condition of the hash values of the same component. The technical framework for performing residual value evaluation on the mobile phone is often suitable for other mobile terminals, firstly, data acquisition and preprocessing are performed, namely, multi-mode data including image data, hardware and configuration data, market price data and the like are acquired, then, feature extraction is performed on the multi-mode data, such as image feature extraction based on a visual model, structural feature extraction through Embedding, abrasion feature extraction by adopting a rule or a light-weight model and the like, then, an old machine depreciation rule is migrated to a new machine through a migration learning module to output uniform equipment feature vectors, then, the fused multi-mode features are input into a residual value prediction model, such as a long-short-term memory network, a Transformer neural network, an RFR (Random Forest Regression, a random forest regression model), an LM-BP network (Levenberg-Marquardt Backpropagation Network, levenberg-Marquardt back propagation neural network) and the like, corresponding residual values, a degree of formation and the like are output, finally, confidence and wind control management are required, and residual value prediction is performed by using confidence degree. The above technology has at least the following technical problems: In the prior art, when an old machine depreciation rule is migrated to a new machine by adopting a migration learning model, the reliability of output migration characteristic data is reduced due to possible dispersion and deletion of sample data input to the migration learning model, so that the data input to a residual value evaluation model is inaccurate, the reliability of an output residual value evaluation result is reduced, and the problem of reduced accuracy of the result of residual value deduction analysis on a mobile terminal exists. Disclosure of Invention In order to solve the technical problem that the accuracy of the pre-deduction result is reduced due to the fact that the reliability of the residual value evaluation result is reduced in the prior art, the embodiment of the invention provides a multi-mode residual value evaluation method and system for pre-deduction of a mobile terminal. The technical proposal is as follows: The method comprises the steps of inputting static data covering functional attributes of a mobile terminal to be evaluated into a migration learning model, analyzing corresponding movable characteristics representing the breakage degree of the mobile terminal to be evaluated after a preset lease, synchronously evaluating the reliability of the movable characteristics, determining whether corresponding reliability adjustment measures are adopted or not based on corresponding evaluation results, wherein the reliability adjustment measures comprise image characteristic deficiency compensation measures, hardware performance characteristic correction measures and abrasion characteristic corre