On-chain pricing method of machine learning valuation model for non-standard asset cash flow prediction

Abstract

The invention relates to the technical field of block chain technology and financial science and technology, in particular to an on-chain pricing method of a machine learning valuation model for non-standard asset cash flow prediction, comprising the following steps: step 1, collecting and preprocessing multi-source heterogeneous data, and obtaining and processing structured data, unstructured text, image data and Internet of Things data; 2, constructing a dynamic integrated machine learning prediction model, including a basic model library, an adaptive weighted integration mechanism and continuous learning optimization; compared with a traditional DCF model, the cash flow prediction accuracy and all pricing data and process upper chain storage are improved through a dynamic integrated learning framework and combined with multi-source heterogeneous data, full-process transparency is achieved, the audit requirement of supervision on financial asset valuation is met, the pricing period is shortened from several weeks to the minute level compared with a manual evaluation method, and the economic benefit is improved. And the non-standard asset transaction efficiency is greatly improved.

Description

Technical Field

[0001] This invention relates to the fields of blockchain technology and fintech, specifically to an on-chain pricing method for a machine learning valuation model for predicting cash flow of non-standard assets. Background Technology

[0002] Non-standard assets face significant challenges from traditional valuation methods due to their unique, non-standardized characteristics and lack of publicly available market transaction data. Existing discounted cash flow (DCF) models rely on subjective assumptions, market comparison methods are limited by the scarcity of comparable transaction cases, and expert evaluation methods suffer from human bias and inefficiency.

[0003] While blockchain technology offers a solution for asset digitization, existing on-chain pricing mechanisms are mostly applicable to standardized tokens or crypto assets, failing to effectively handle the complex cash flow characteristics and risk attributes of non-standard assets. Furthermore, traditional machine learning models perform poorly when dealing with the high-dimensional sparse data, nonlinear relationships, and unstructured information of non-standard assets, and their interpretability is insufficient, making it difficult to meet financial regulatory requirements. Therefore, this paper proposes an on-chain pricing method for a machine learning valuation model of non-standard asset cash flow forecasting. Summary of the Invention

[0004] In view of this, the present invention provides an on-chain pricing method for a machine learning valuation model for non-standard asset cash flow forecasting, in order to solve or alleviate the technical problems existing in the prior art, and at least provide a beneficial alternative.

[0005] The technical solution of this invention is implemented as follows: an on-chain pricing method for a machine learning valuation model of non-standard asset cash flow forecasting, comprising the following steps:

[0006] Step 1: Multi-source heterogeneous data acquisition and preprocessing, acquiring and processing structured data, unstructured text, image data and IoT data;

[0007] Step 2: Dynamically integrate machine learning prediction model construction, including basic model library, adaptive weighted ensemble mechanism and continuous learning optimization;

[0008] Step 3: On-chain pricing smart contract execution. On-chain pricing is executed through a smart contract, including recording pricing parameters on-chain, smart contract pricing logic, and multi-signature verification mechanism.

[0009] Step 4: Enhanced Explainability and Regulatory Compliance. Enhance model explainability and ensure regulatory compliance, including generating explainability reports and audit trails.

[0010] More preferably, in step one, asset ownership records, historical transaction data, and mortgage information are obtained from blockchain nodes; macroeconomic indicators (interest rates, inflation rates) and industry data are obtained from financial databases; asset-related legal texts (contract terms, lease agreements), news reports, and appraisal reports are analyzed using natural language processing (NLP) technology; physical state images of assets (property appearance, infrastructure conditions) are processed using computer vision (CV) technology; asset operation data (such as tenant occupancy rates and energy consumption data for commercial real estate) are collected in real time; missing values ​​in structured data are filled based on time series interpolation algorithms (such as Kalman filtering); missing fields in unstructured text are supplemented using named entity recognition (NER) technology; the Isolation Forest algorithm is applied to detect and correct abnormal price fluctuations in transaction data; and a multi-dimensional feature set is constructed, including inherent asset characteristics (geographical location, type), time series characteristics (historical cash flow trends), textual characteristics (legal clause risk scores), and image characteristics (building quality scores).

[0011] In a further preferred embodiment, in step two, the weights of each basic model are dynamically adjusted based on the meta-learning algorithm. The weight update frequency is automatically determined according to the data volatility. A model confidence index (calculated based on the prediction interval width and historical accuracy) is introduced. When the confidence of a single model is lower than a threshold (e.g., 0.6), a model fusion mechanism is triggered. An online learning framework is adopted to update the model parameters in real time using newly incoming data, while retaining historical knowledge. A concept drift detection mechanism is designed to automatically trigger model structure adjustment when a significant change in data distribution is detected.

[0012] In a further preferred embodiment, in step three, the preprocessed data features, model weights, and prediction results are compressed using a Merkle tree structure and stored on the blockchain to ensure data integrity and traceability. Zero-knowledge proof (ZKP) technology is used to verify the authenticity of the data source, protecting sensitive information from leakage. Based on the predicted future cash flow, the present value of the asset is calculated using a risk-adjusted discount rate. Transaction data and social media sentiment indices on the blockchain are introduced to dynamically adjust the pricing results. Based on the historical liquidity data of the asset on the on-chain trading platform, a liquidity risk premium is calculated. A multi-party pricing verification process is designed, requiring at least N authorized nodes (such as appraisal institutions, regulatory authorities, and asset management companies) to reach a consensus (more than 2 / 3 agreement) before the final price can be confirmed.

[0013] In a further preferred embodiment, in step four, the contribution of each feature to the pricing result is quantified using SHAP (SHapley Additive ex Planations) value technology, a visual explanation report is generated, a legal compliance checker is developed, and the pricing process is automatically verified to ensure compliance with financial regulatory requirements (such as Basel III and IFRS 13). The entire pricing process (data collection timestamps, model parameter change records, and verification signatures) is recorded on the blockchain to form an immutable audit trail, and an API interface is provided for regulatory agencies to query and retrieve pricing-related data in real time.

[0014] The embodiments of the present invention have the following advantages due to the adoption of the above technical solutions:

[0015] I. This invention improves the accuracy of cash flow forecasting compared to traditional DCF models by using a dynamic integrated learning framework combined with multi-source heterogeneous data. All pricing data and processes are stored on the blockchain, achieving full transparency and meeting regulatory audit requirements for financial asset valuation. Compared to manual assessment methods, the pricing cycle is shortened from weeks to minutes, significantly improving the efficiency of non-standard asset transactions.

[0016] Second, this invention captures market changes in a timely manner through a continuous learning mechanism and market sentiment correction, reduces the risk caused by valuation deviations, provides explanations of feature importance and automatic compliance checks, solves the interpretability problem of traditional black-box models, and meets regulatory requirements for model transparency.

[0017] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0020] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0021] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] like Figure 1 As shown, this embodiment of the invention provides an on-chain pricing method for a machine learning valuation model of non-standard asset cash flow forecasting, including the following steps:

[0023] Step 1: Multi-source heterogeneous data acquisition and preprocessing, acquiring and processing structured data, unstructured text, image data and IoT data;

[0024] Step 2: Dynamically integrate machine learning prediction model construction, including basic model library, adaptive weighted ensemble mechanism and continuous learning optimization;

[0025] Step 3: On-chain pricing smart contract execution. On-chain pricing is executed through a smart contract, including recording pricing parameters on-chain, smart contract pricing logic, and multi-signature verification mechanism.

[0026] Step 4: Enhanced Explainability and Regulatory Compliance. Enhance model explainability and ensure regulatory compliance, including generating explainability reports and audit trails.

[0027] In one embodiment, in step one, asset ownership records, historical transaction data, and mortgage information are obtained from blockchain nodes; macroeconomic indicators (interest rates, inflation rates), industry data, etc., are obtained from financial databases; asset-related legal texts (contract terms, lease agreements), news reports, and appraisal reports are parsed using natural language processing (NLP) technology; physical state images of assets (property appearance, infrastructure conditions) are processed using computer vision (CV) technology; asset operation data (such as tenant occupancy rates and energy consumption data for commercial real estate) are collected in real time; missing values ​​in structured data are filled based on time series interpolation algorithms (such as Kalman filtering); missing fields in unstructured text are supplemented using named entity recognition (NER) technology; the Isolation Forest algorithm is applied to detect and correct abnormal price fluctuations in transaction data; and a multi-dimensional feature set is constructed, including inherent asset characteristics (geographical location, type), time series characteristics (historical cash flow trends), text features (legal clause risk scores), and image features (building quality scores).

[0028] In one embodiment, in step two, the weights of each base model are dynamically adjusted based on the meta-learning algorithm. The weight update frequency is automatically determined according to the data volatility. A model confidence index (calculated based on the prediction interval width and historical accuracy) is introduced. When the confidence of a single model is lower than a threshold (e.g., 0.6), a model fusion mechanism is triggered. An online learning framework is adopted to update the model parameters in real time using newly incoming data, while retaining historical knowledge. A concept drift detection mechanism is designed to automatically trigger model structure adjustment when a significant change in data distribution is detected.

[0029] In one embodiment, in step three, the preprocessed data features, model weights, prediction results, etc., are compressed using a Merkle tree structure and stored on the blockchain to ensure data integrity and traceability. Zero-knowledge proof (ZKP) technology is used to verify the authenticity of the data source and protect sensitive information from leakage. Based on the predicted future cash flow, the present value of the asset is calculated using a risk-adjusted discount rate. Transaction data and social media sentiment indices on the blockchain are introduced to dynamically adjust the pricing results. Based on the historical liquidity data of the asset on the on-chain trading platform, a liquidity risk premium is calculated. A pricing verification process involving multiple parties is designed, requiring at least N authorized nodes (such as appraisal institutions, regulatory authorities, and asset management companies) to reach a consensus (more than 2 / 3 agree) before the final price can be confirmed.

[0030] In one embodiment, in step four, the contribution of each feature to the pricing result is quantified using SHAP (SHapley Additive ex Planations) value technology, a visual explanation report is generated, a legal compliance checker is developed, and the pricing process is automatically verified to ensure compliance with financial regulatory requirements (such as Basel III and IFRS 13). The entire pricing process (data collection timestamps, model parameter change records, and verification signatures) is recorded on the blockchain to form an immutable audit trail, and an API interface is provided for regulatory agencies to query and retrieve pricing-related data in real time.

[0031] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in the present invention, and these should all be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An on-chain pricing method for a machine learning valuation model of non-standard asset cash flow forecasting, characterized by: Includes the following steps: Step 1: Multi-source heterogeneous data acquisition and preprocessing, acquiring and processing structured data, unstructured text, image data and IoT data; Step 2: Dynamically integrate machine learning prediction model construction, including basic model library, adaptive weighted ensemble mechanism and continuous learning optimization; Step 3: On-chain pricing smart contract execution. On-chain pricing is executed through a smart contract, including recording pricing parameters on-chain, smart contract pricing logic, and multi-signature verification mechanism. Step 4: Enhanced Explainability and Regulatory Compliance. Enhance model explainability and ensure regulatory compliance, including generating explainability reports and audit trails.

2. The on-chain pricing method for the machine learning valuation model of non-standard asset cash flow forecasting according to claim 1, characterized in that: In step one, asset ownership records, historical transaction data, and mortgage information are obtained from blockchain nodes; macroeconomic indicators (interest rates, inflation rates) and industry data are obtained from financial databases; natural language processing (NLP) technology is used to parse asset-related legal texts (contract terms, lease agreements), news reports, and appraisal reports; computer vision (CV) technology is used to process images of the physical state of the assets (property appearance, infrastructure condition); real-time asset operation data (such as tenant occupancy rates and energy consumption data for commercial real estate) is collected; and missing values ​​in structured data are filled using time series interpolation algorithms (such as Kalman filtering); missing fields in unstructured text are supplemented using named entity recognition (NER) technology; the Isolation Forest algorithm is applied to detect and correct abnormal price fluctuations in transaction data; and a multi-dimensional feature set is constructed, including inherent asset characteristics (geographical location, type), time series characteristics (historical cash flow trends), textual characteristics (legal clause risk scores), and image characteristics (building quality scores).

3. The on-chain pricing method for the machine learning valuation model of non-standard asset cash flow forecasting according to claim 1, characterized in that: In step two, the weights of each base model are dynamically adjusted based on the meta-learning algorithm. The weight update frequency is automatically determined according to the data volatility. A model confidence index (calculated based on the prediction interval width and historical accuracy) is introduced. When the confidence of a single model is lower than a threshold (e.g., 0.6), a model fusion mechanism is triggered. An online learning framework is adopted to update the model parameters in real time using newly incoming data, while retaining historical knowledge. A concept drift detection mechanism is designed to automatically trigger model structure adjustment when a significant change in data distribution is detected.

4. The on-chain pricing method for the machine learning valuation model of non-standard asset cash flow forecasting according to claim 1, characterized in that: In step three, the preprocessed data features, model weights, and prediction results are compressed using a Merkle tree structure and stored on the blockchain to ensure data integrity and traceability. Zero-knowledge proof (ZKP) technology is used to verify the authenticity of the data source and protect sensitive information from leakage. Based on the predicted future cash flow, the present value of the asset is calculated using a risk-adjusted discount rate. Transaction data and social media sentiment indices on the blockchain are introduced to dynamically adjust the pricing results. Based on the historical liquidity data of the asset on the on-chain trading platform, a liquidity risk premium is calculated. A multi-party pricing verification process is designed, requiring at least N authorized nodes (such as appraisal institutions, regulatory authorities, and asset management companies) to reach a consensus (more than 2 / 3 agreement) before the final price can be confirmed.

5. The on-chain pricing method for the machine learning valuation model of non-standard asset cash flow forecasting according to claim 1, characterized in that: In step four, the contribution of each feature to the pricing result is quantified using SHAP (SHapley Additive ex Planations) value technology, a visual explanation report is generated, a legal compliance checker is developed, and the pricing process is automatically verified to ensure compliance with financial regulatory requirements (such as Basel III and IFRS 13). The entire pricing process (data collection timestamps, model parameter change records, and verification signatures) is recorded on the blockchain to form an immutable audit trail. An API interface is provided for regulatory agencies to query and retrieve pricing-related data in real time.

Images