Mineral rights transaction large model response system based on weighted data modeling

Abstract

The present invention discloses a large model response system for mining rights transactions based on weighted data modeling, comprising the following steps: receiving and identifying input instructions based on a large language model, analyzing query intent, the query intent including historical data query and mining rights value estimation; if the query intent is historical data query, directly searching the database and returning the result; if the query intent is mining rights value estimation, inputting the received mine data into the weighted data model, combining the machine learning model calculation result, generating the estimated mining rights transaction value and returning the result, the system of the present invention combines the retrieval enhancement generation framework through the large language model, collects and analyzes multi-dimensional data, uses the machine learning algorithm to set reasonable weights for different parameters, and constructs the complex relationship between the mining rights transaction value and these parameters. The user can query the historical transaction data in the form of natural language dialogue, or input the specific information of the current mine, and the system can automatically output the mining rights transaction value.

Description

Technical Field

[0001] The present invention relates to the technical field of mining rights trading, and in particular to a mining rights trading large model response system based on weighted data modeling. Background Art

[0002] With the increasing importance of global mineral resources and the continuous expansion of the mining rights trading market, the accurate assessment of the value of mining rights transactions has become a common focus of mining companies, investors and government agencies. Traditional methods of evaluating the value of mining rights transactions rely on expert experience, historical transaction cases and simple statistical models. These methods often have problems such as strong subjectivity, poor timeliness and insufficient accuracy, and are difficult to meet the current complex and changing mining market needs. In addition, the mathematical model established by mathematical formulas is not readable, and more often than not, users do not understand the formulas and cannot provide corresponding values ​​according to the parameters. They can only explain their needs in verbal terms, and conventional programs cannot understand the user's intentions.

[0003] In recent years, the rapid development of big data, artificial intelligence and natural language processing technology has provided new possibilities for the intelligent evaluation of the value of mining rights transactions. As a core technology in the field of natural language processing, Large Language Model (LLM) has been widely used in many industries with its powerful language generation and understanding capabilities, significantly improving the efficiency and accuracy of information processing and decision-making. However, when LLM is directly applied to the highly professional and data-intensive field of mining rights transactions, it faces problems such as insufficient professional knowledge, data diversity and non-stationarity, which limits its full performance. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention provides a large model response system for mining rights transactions based on weighted data modeling. The system collects and analyzes multi-dimensional data through a large language model combined with a retrieval enhancement generation framework, and uses machine learning algorithms to set reasonable weights for different parameters to construct a complex relationship between the value of mining rights transactions and these parameters. Users can query historical transaction data through natural language dialogues or input specific information about the current mine. The system can automatically output the estimated value of mining rights transactions, providing efficient and accurate decision support for mining companies, investors and government agencies.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] The present invention provides a large-scale model response system for mining rights transactions based on weighted data modeling, comprising the following steps:

[0007] Receive and recognize input commands based on a large language model, and analyze query intent, including historical data query and mining rights value estimation;

[0008] If the query is intended for historical data, the database is directly searched and the results are returned;

[0009] If the query intent is to estimate the value of mining rights, the received mining data will be input into the weighted data model, combined with the calculation results of the machine learning model to generate the estimated mining rights transaction value and return the result.

[0010] In the present invention, the user's inquiries are no longer answered based on a simple large language model. Instead, the user's query intention is analyzed based on the large language model. In simple inquiries, the database result answer can be directly called. When consulting on the value of mining rights transaction, the retrieval enhancement generation framework is called, that is, the reasonable weights set for different parameters by using machine learning algorithms, to construct a mathematical model that can accurately reflect the complex relationship between the mining rights transaction value and each parameter, and calculate a highly accurate mining rights transaction value answer.

[0011] The system of the present invention collects and analyzes multi-dimensional data through a large language model combined with a search enhancement generation framework, uses machine learning algorithms to set reasonable weights for different parameters, and constructs a complex relationship between the value of mining rights transactions and these parameters. Users can query historical transaction data through natural language dialogues or input specific information about the current mine. The system can automatically output the estimated value of mining rights transactions, providing efficient and accurate decision support for mining companies, investors, and government agencies.

[0012] In a further technical solution, the mine data includes mine area, mineral material type and service life.

[0013] All parameters that have a huge impact on the transaction value of mining rights are collected to ensure the accuracy of the calculation results.

[0014] In a further technical solution, a large language model is combined with a database when analyzing query intent.

[0015] When analyzing query intent, combining data keywords in the database can more quickly and accurately assist the large language model in semantic recognition and keyword recognition, thereby improving efficiency and accuracy.

[0016] In a further technical solution, the weighted data model is:

[0017] ;

[0018] Among them, V is the transaction value of mining rights, A is the area of ​​the mine, is the weight of the mine area, T is the unit price of mineral materials, is the unit price weight of mineral materials, Y is the useful life, is the useful life weight.

[0019] The transaction value of mining rights that takes multiple weights into consideration has higher accuracy.

[0020] In a further technical solution, the weight calculation method includes the following steps:

[0021] First, the area weight is calculated. By retrieving the data with the same unit price of mineral materials in the data set, the data is grouped according to the service life, so as to construct a 1-year group, a 2-year group, and n-year group. The variance of the mine area weight of each group of data is calculated to obtain the weight data of the same mineral materials with different service lives. The variance formula is:

[0022] ,in, is the mining area of ​​each group of data in the 1-year group, is the mining rights transaction value corresponding to the mining area of ​​each group in the 1-year group, It is an unknown number. By floating up and down the preset value, we can find The minimum value of ;

[0023] Then calculate the 2-year group in the same way. , 3-year group ...n-year group ; Then the 1-year group , 2nd year group , 3-year group ...n-year group The value is weighted averaged through each group of parameters to obtain the final unique ;

[0024] exist After determination, the same method is used to calculate and That's it.

[0025] For the weight calculation of different parameters, the weight with minimum variance is adopted, which has higher accuracy.

[0026] In a further technical solution, when identifying the input instruction, if the input is the coordinates of the inflection point of the mine, the mine area is calculated according to the Gaussian area formula, and then the subsequent mining rights transaction value calculation is performed.

[0027] Since the age of the mine and the unit price of the ore can be directly obtained from the instructions, and ordinary people are often unable to directly provide the accurate mine area, the mine area can be obtained through calculation to further improve the accuracy.

[0028] In a further technical solution, the establishment of the machine learning model includes the following steps:

[0029] Preprocess the database data, including data cleaning, data standardization and data partitioning;

[0030] Select features that have a significant impact on the value of mining rights transactions from the preprocessed data set, and use feature extraction technology to convert text data into numerical features;

[0031] Based on the selected feature set, the model is trained using a machine learning algorithm; during the training process, the model parameters are adjusted through cross-validation and grid search methods.

[0032] Through precise data processing, subsequent model training is facilitated, and the predictive performance of the model is ensured in a variety of ways, further improving the accuracy of the value of mining rights transactions.

[0033] In a further technical solution, calculating the mine area includes the following steps:

[0034] Arrange the received coordinate points into a list and get ( )、( )……( ) of a series of mine coordinates, through the Gaussian area formula:

[0035] ;

[0036] Calculate the mine area.

[0037] Through the Gaussian area formula and the calculation of coordinate points, the accurate mine area can be quickly obtained. At the same time, the form of only inputting coordinate points requires lower professional ability for ordinary users. It only requires natural language dialogue or manual input of coordinate point data.

[0038] The beneficial effects are:

[0039] 1. The system of the present invention, by introducing the interactive mode of natural language dialogue, users can quickly obtain the required historical data or estimated value of mining rights transactions without cumbersome query operations, which greatly improves the convenience and efficiency of information acquisition and the convenience of obtaining mining rights transaction information; by adopting weighted data modeling and machine learning algorithms, the system can automatically learn and optimize the complex relationship between parameters such as mine area, mineral material type, service life and mine amount, so as to provide more accurate and reliable mining rights transaction value estimation, provide strong support for decision makers, and enhance the accuracy of mining rights transaction value assessment. The system provides real-time and accurate mining rights transaction value estimation. It helps mining companies and related institutions to better evaluate the market value of mining resources, optimize resource allocation, and improve the scientificity and efficiency of decision-making. The automated and intelligent response system reduces the workload of manual query and evaluation and reduces labor costs. At the same time, the instant response feature also shortens the decision-making cycle and reduces time costs. The system provides public and verifiable historical data and estimated values ​​of mining rights transactions, which helps increase market transparency, reduce information asymmetry, and promote the standardized development of the mining rights trading market. Accurate value assessment helps mining companies to rationally plan mining plans, avoid over-exploitation and waste of resources, and promote the sustainable development of the mining industry.

[0040] 2. All parameters that have a huge impact on the transaction value of mining rights are collected to ensure the accuracy of the calculation results.

[0041] 3. When analyzing query intent, combining data keywords in the database can more quickly and accurately assist the large language model in semantic recognition and keyword recognition, thereby improving efficiency and accuracy.

[0042] 4. The transaction value of mining rights with multiple weights considered has higher accuracy.

[0043] 5. For the weight calculation of different parameters, the weight with minimum variance is adopted, which has higher accuracy.

[0044] 6. Since the age of the mine and the unit price of the ore can be directly obtained from the instructions, and ordinary people are often unable to directly provide the accurate mine area, the mine area can be obtained through calculation to further improve the accuracy.

[0045] 7. Through precise data processing, subsequent model training is facilitated, and the prediction performance of the model is ensured in a variety of ways, further improving the accuracy of the value of mining rights transactions.

[0046] 8. Through the Gaussian area formula and the calculation of coordinate points, the accurate mine area can be quickly obtained. At the same time, the form of only inputting coordinate points requires lower professional ability for ordinary users. It only requires natural language dialogue or manual input of coordinate point data. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a logical flow diagram of a large model response system for mining rights transactions based on weighted data modeling in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The present invention will be further described below in conjunction with the accompanying drawings:

[0049] Example:

[0050] like Figure 1 As shown, a large model response system for mining rights transactions based on weighted data modeling includes the following steps:

[0051] Receive and recognize input commands based on a large language model, and analyze query intent, including historical data query and mining rights value estimation;

[0052] Specifically, a user's query is received as an input instruction, keywords in the query are extracted, and the user's intention is identified.

[0053] If the query is intended for historical data, the database is directly searched and the results are returned;

[0054] Specifically, without involving the intention of the algorithm model, the existing RAG framework is used to retrieve documents or information in the relevant knowledge base;

[0055] The RAG framework is a retrieval augmented generation framework that achieves accurate answers to complex questions by combining an external knowledge base with a large language model (LLM).

[0056] If the query intent is to estimate the value of mining rights, the received mining data will be input into the weighted data model, combined with the calculation results of the machine learning model to generate the estimated mining rights transaction value and return the result.

[0057] In the present invention, the user's inquiries are no longer answered based on a simple large language model. Instead, the user's query intention is analyzed based on the large language model. In simple inquiries, the database result answer can be directly called. When consulting on the value of mining rights transaction, the retrieval enhancement generation framework is called, that is, the reasonable weights set for different parameters by using machine learning algorithms, to construct a mathematical model that can accurately reflect the complex relationship between the mining rights transaction value and each parameter, and calculate a highly accurate mining rights transaction value answer.

[0058] The system of the present invention collects and analyzes multi-dimensional data through a large language model combined with a search enhancement generation framework, uses machine learning algorithms to set reasonable weights for different parameters, and constructs a complex relationship between the value of mining rights transactions and these parameters. Users can query historical transaction data through natural language dialogues or input specific information about the current mine. The system can automatically output the estimated value of mining rights transactions, providing efficient and accurate decision support for mining companies, investors, and government agencies.

[0059] In another embodiment, the mine data includes mine area, mineral material type and service life.

[0060] All parameters that have a huge impact on the transaction value of mining rights are collected to ensure the accuracy of the calculation results.

[0061] In another embodiment, when analyzing query intent, a large language model is combined with a database.

[0062] When analyzing query intent, combining data keywords in the database can more quickly and accurately assist the large language model in semantic recognition and keyword recognition, thereby improving efficiency and accuracy.

[0063] In another embodiment, the weighted data model is:

[0064] ;

[0065] Among them, V is the transaction value of mining rights, A is the area of ​​the mine, is the weight of the mine area, T is the unit price of mineral materials, is the unit price weight of mineral materials, Y is the useful life, is the useful life weight.

[0066] The transaction value of mining rights that takes multiple weights into consideration has higher accuracy.

[0067] In another embodiment, the weight calculation method includes the following steps:

[0068] First, the area weight is calculated. By retrieving the data with the same unit price of mineral materials in the data set, the data is grouped according to the service life, so as to construct a 1-year group, a 2-year group, and n-year group. The variance of the mine area weight of each group of data is calculated to obtain the weight data of the same mineral materials with different service lives. The variance formula is:

[0069] ,in, is the mining area of ​​each group of data in the 1-year group, is the mining rights transaction value corresponding to the mining area of ​​each group in the 1-year group, It is an unknown number. By floating up and down the preset value, we can find The minimum value of ;

[0070] Then calculate the 2-year group in the same way. , 3-year group ...n-year group ; Then the 1-year group , 2nd year group , 3-year group ...n-year group The value is weighted averaged through each group of parameters to obtain the final unique ;

[0071] exist After determination, the same method is used to calculate and That's it.

[0072] For the weight calculation of different parameters, the weight with minimum variance is adopted, which has higher accuracy.

[0073] In another embodiment, when the input instruction is recognized, if the input is the coordinates of the inflection point of a mine, the area of ​​the mine is calculated according to the Gaussian area formula, and then the subsequent mining rights transaction value calculation is performed.

[0074] It can be understood that the mine inflection point coordinates referred to in this embodiment are national geodetic coordinates.

[0075] Since the age of the mine and the unit price of the ore can be directly obtained from the instructions, and ordinary people are often unable to directly provide the accurate mine area, the mine area can be obtained through calculation to further improve the accuracy.

[0076] In another embodiment, the establishment of the machine learning model includes the following steps:

[0077] Preprocess the database data, including data cleaning (removing noise, outliers, etc.), data standardization (unifying units, formats, etc.) and data partitioning (training set, validation set, test set, etc.);

[0078] Select features that have a significant impact on the value of mining rights transactions (such as mine area, mineral material type, service life, etc.) from the preprocessed data set, and use feature extraction technology (such as word embedding, autoencoder, etc.) to convert text data (such as mineral material type description) into numerical features;

[0079] Based on the selected feature set, machine learning algorithms (such as linear regression, random forest, gradient boosting tree, etc.) are used to train the model. During the training process, model parameters (including weight distribution) are adjusted through cross-validation and grid search methods to optimize the predictive performance of the model. At the same time, regularization, feature selection and other technical means are used to prevent the occurrence of overfitting.

[0080] Through precise data processing, subsequent model training is facilitated, and the predictive performance of the model is ensured in a variety of ways, further improving the accuracy of the value of mining rights transactions.

[0081] In another embodiment, calculating the mine area includes the following steps:

[0082] Arrange the received coordinate points into a list and get ( )、( )……( ) of a series of mine coordinates, through the Gauss area formula (also known as the Shoelace formula):

[0083] ;

[0084] The mine area is calculated by connecting the first vertex with the last vertex to form a closed polygon, calculating the product of each vertex with the next vertex, and finally adding up the products to get the mine area.

[0085] It should be emphasized that the coordinate points referred to in this embodiment are national geodetic coordinates.

[0086] Through the Gaussian area formula and the calculation of coordinate points, the accurate mine area can be quickly obtained. At the same time, the form of only inputting coordinate points requires lower professional ability for ordinary users. It only requires natural language dialogue or manual input of coordinate point data.

[0087] It is understandable that this is not a basic calculation improvement, but after combining with the large language model, it fully considers the professional ability of the user and the adequacy of the basic conditions. Often, the basic data only contains the mine coordinates but not the precise area. Therefore, this embodiment can satisfy the simple and convenient input of coordinates by ordinary users through natural language input and calculation of the calculation module, and the system quickly calculates the area parameters for subsequent calculation of the value of the mining rights, which improves the efficiency and accuracy of the overall calculation.

[0088] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A large-scale model response system for mining rights transactions based on weighted data modeling, characterized in that: The following steps are involved: Receive and recognize input commands based on a large language model, and analyze query intent, including historical data query and mining rights value estimation; If the query is intended for historical data, the database is directly searched and the results are returned; If the query intent is to estimate the value of mining rights, the received mining data is input into the weighted data model, combined with the calculation results of the machine learning model, to generate the estimated mining rights transaction value and return the result; The weighted data model is: Among them, V is the transaction value of mining rights, A is the mine area, μA is the mine area weight, T is the unit price of mineral materials, μT is the unit price weight of mineral materials, Y is the useful life, and μY is the useful life weight; The weight calculation method includes the following steps: First, the area weight is calculated. By retrieving the data with the same unit price of mineral materials in the data set, the data is grouped according to the service life, so as to construct a 1-year group, a 2-year group, and n-year group. The variance of the mine area weight of each group of data is calculated to obtain the weight data of the same mineral materials with different service lives. The variance formula is: Among them, A1, A2, ... A n is the mining area of ​​each group of data in the 1-year group, V1, V2, ... V n is the mining rights transaction value corresponding to the mining area of ​​each group in the 1-year group. μA1 is an unknown number. By floating up and down the preset value, find the minimum value of x to determine μA1. Then, calculate μA2 of the 2-year group, μA3 of the 3-year group, and μA of the n-year group in the same way. n ; Then add μA1 of the 1-year group, μA2 of the 2-year group, μA3 of the 3-year group, and so on μA of the n-year group n The value is weighted averaged through each group of parameter quantities to obtain the final unique μA; After μA is determined, μT and μY can be calculated using the same method.

2. The mining rights transaction large model response system based on weighted data modeling according to claim 1 is characterized in that: The mine data include mine area, mineral material types and service life.

3. The mining rights transaction large model response system based on weighted data modeling according to claim 1 is characterized in that: When analyzing query intent, a large language model is combined with the database.

4. The mining rights transaction large model response system based on weighted data modeling according to claim 2 is characterized in that: When recognizing the input command, if the input is the coordinates of the mine inflection point, the mine area is calculated according to the Gaussian area formula, and then the subsequent mining rights transaction value calculation is performed.

5. The mining rights transaction large model response system based on weighted data modeling according to claim 1 is characterized in that: The establishment of the machine learning model includes the following steps: Preprocess the database data, including data cleaning, data standardization and data partitioning; Select features that have a significant impact on the value of mining rights transactions from the preprocessed data set, and use feature extraction technology to convert text data into numerical features; Based on the selected feature set, the model is trained using a machine learning algorithm; during the training process, the model parameters are adjusted through cross-validation and grid search methods.

6. The mining rights transaction large model response system based on weighted data modeling according to claim 4 is characterized in that: Calculating the mine area involves the following steps: The received coordinate points are sorted into a list to obtain (x1, y1), (x2, y2)...(x n ,y n ) of a series of mine coordinates, through the Gaussian area formula: Calculate the mine area.

Images