Automatic bond pricing method and system

Abstract

The invention discloses a bond automatic pricing method and system, and the method comprises the steps: operating a bond representation calculation model, and opening a bond vector database; pre-processing the new transaction bond transaction data to obtain pre-processed recent new transaction bond transaction data, and updating the bond representation calculation model and the bond vector database; inputting the preprocessed to-be-predicted bond into a bond representation calculation model to obtain a to-be-predicted bond representation vector; constructing a dynamic comparison library of bonds to be predicted; and inputting the to-be-predicted bond representation vector and the dynamic comparison library into a similar bond calculation module to obtain a similar bond ordered list corresponding to the to-be-predicted bond, and inputting the similar bond ordered list into a pricing module to obtain the price of the to-be-predicted bond. According to the invention, bond automatic pricing and similar bond automatic recommendation based on deep learning are realized, and the accuracy and efficiency of bond pricing are improved. A bond reference price is provided for a dealer or an investment manager without experience, and the intelligence and precision of a bond pricing market are improved.

Description

Technical Field

[0001] The present invention belongs to the field of digital finance, and in particular relates to a bond automatic pricing method and system. Background Art

[0002] The bond market is a place for issuing and buying and selling bonds, and is an important part of the financial market. Compared with the stock market, the bond market is more discrete. The stock price system has a real-time price, but the bond system only gives a reference price for a key period. There are many factors to consider in bond pricing, including China bond valuation (exercise), China bond valuation (maturity), transaction time, remaining term, debt subject and other multi-dimensional factors. These factors make bond pricing face huge challenges.

[0003] Most of the existing methods are that investment managers rely on experience to select the attributes based on which they use as features for pricing analysis, which cannot achieve automated and intelligent pricing. For example, according to the document "Liu Lu, Zhang Junrui. Method and Application of Credit Bond Pricing [J]. Bonds, 2024 (1): 51-56", the core logic of bond pricing, especially credit bond pricing, is to make adjustments based on similar bond interest rates. If the same subject has bonds with the same terms and terms and a reliable market interest rate, then this market interest rate can be directly used as the pricing result, but this ideal state rarely occurs in reality, so some pricing methods are needed. Whether in the primary market or the secondary market, the core logic of bond pricing is the same, that is, similar bonds must be found first, and then the yield to maturity of similar bonds must be adjusted. Although the existing methods conform to this pricing logic, they still have shortcomings. For example, patent CN201910935365.6 "A Bond Pricing Method and System" regards the most recently issued bonds with the same issuance information as the new bond on the day of issuance as similar bonds. Although it can achieve a certain effect, the issuance information is limited to a few main attributes and cannot fully mine similar bonds.

[0004] In addition, there are methods for directly calculating pricing, such as patent CN201010139142.8 "A floating rate bond pricing and risk arbitrage method" which directly uses formulas to calculate bond pricing. For example, patent CN202010281602.4 "A convertible bond value assessment system and method" provides a method for constructing a partial differential equation model to calculate pricing. Although the parameters used in the above two methods are calculated based on stock or bond data over a period of time, the calculation results lack interpretability, and similar bonds still need to be collected in actual use. Summary of the invention

[0005] In view of the above technical problems, the present invention discloses a bond automatic pricing method and system, aiming to solve the problem that similar bonds cannot be automatically found and priced in the prior art.

[0006] The present invention solves the technical problem by adopting the following technical solution: In a first aspect, the present invention provides a bond automatic pricing method, which comprises the following steps: Step S1, running the bond representation calculation model and opening the bond vector database; Step S2, preprocessing the newly traded bond transaction data to obtain the preprocessed recent newly traded bond transaction data, and updating the bond representation calculation model and bond vector database; Step S3, inputting the preprocessed bond to be predicted into the bond representation calculation model to obtain a bond representation vector to be predicted; Step S4, constructing a dynamic comparison library of bonds to be predicted; Step S5, input the representation vector of the bond to be predicted and the dynamic comparison library into the similar bond calculation module, obtain an ordered list of similar bonds corresponding to the bond to be predicted, input it into the pricing module, and obtain the pricing of the bond to be predicted.

[0007] In one implementation, in step S1, the training process of the bond representation calculation model includes: Acquire historical bond trading data from the bond market, and preprocess the historical bond trading data to obtain preprocessed historical bond trading data; The pre-processed historical bond trading data is input into a deep learning neural network for training, and a bond representation calculation model is obtained after the training is completed.

[0008] In one implementation, in step S1, the process of constructing the bond vector database includes: Inputting the preprocessed historical bond transaction data into the bond representation calculation model to obtain a historical bond representation vector; The historical bond transaction data and the historical bond representation vectors are associated and indexed to construct a bond vector database.

[0009] In one implementation, in step S2, the bond transaction data is time point data, and each piece of data is an attribute of a bond when it is traded at a time point; Step S2 needs to be performed regularly at regular intervals.

[0010] The preprocessing steps include data cleaning, converting time attributes into numerical attributes, converting Boolean type attributes into numerical attributes, sequentially encoding category type attributes, and normalizing all numerical attributes and sequential encodings.

[0011] In one implementation, in step S4, constructing a dynamic comparison library of the bonds to be predicted includes screening out bond trading data whose transaction time is before the pricing time of the bonds to be predicted and their corresponding bond representation vectors from the bond vector database to construct a dynamic comparison library.

[0012] In one implementation, in step S5, the step of inputting the bond representation vector to be predicted and the dynamic comparison library into the similar bond calculation module to obtain an ordered list of similar bonds corresponding to the bond to be predicted includes: Set the number of similar bonds N; Calculating the vector similarity between the bond representation vector in the dynamic comparison library and the bond representation vector to be predicted, and sorting them; The top N bond transaction data with the highest similarity are selected to form an ordered list of similar bonds of the bond to be predicted.

[0013] In a second aspect, the present invention further provides a bond automatic pricing system, comprising: The bond preprocessing module is used to preprocess the attributes of each bond transaction data, including data cleaning, converting time attributes into numerical attributes, converting Boolean type attributes into numerical attributes, sequentially encoding category type attributes, and normalizing all numerical attributes and sequential encodings; A bond encoding module is used to encode the preprocessed bond trading data, including the preprocessed bond trading data to be predicted and the preprocessed historical bond trading data, and input the preprocessed bond trading data into the bond representation calculation model to obtain the bond representation vector to be predicted and the historical bond representation vector for characterizing the attribute information of the bond; A bond vector database construction module associates the historical bond transaction data with the historical bond representation vector and creates an index to construct a bond vector database for fast retrieval of bond transaction data and representation vectors; A comparison library construction module, used to obtain the pricing time of the bond to be predicted, and construct a dynamic comparison library of the bond to be predicted from the bond vector database according to the pricing time; A similar bond calculation module is used to calculate the similarity between the bond representation vector to be predicted and the bond representation vector in the dynamic comparison library, and the similarity is used to characterize the similarity between the bond to be predicted and the historical bond transaction data; based on the similarity, the bond transaction data in the dynamic comparison library are sorted to obtain a number of bond transaction data with the highest similarity to form an ordered list of similar bonds; A pricing module, used for predicting the bond yield to be predicted based on the bond yields in the ordered list of similar bonds; The update module is used to regularly obtain the newly traded bond transaction data in the recent bond market, pre-process the newly traded bond transaction data to obtain the pre-processed recent newly traded bond transaction data, and update the bond representation calculation model and the bond vector database.

[0014] In one implementation, the bond preprocessing module includes: Data cleaning unit, used to remove duplicate data; Attribute filling unit, used to complete the missing attributes of bonds; A coding unit, used to convert a time attribute into a numerical attribute, convert a Boolean type attribute into a numerical attribute, and perform sequential coding on a category type attribute; Normalization unit, used to normalize all numeric attributes and ordinal encodings.

[0015] In one implementation, the similar coupon calculation module includes: A similarity calculation unit, used to calculate the similarity between the bond representation vector to be predicted and the bond representation vector in the dynamic comparison library, wherein the similarity is used to characterize the similarity between the bond to be predicted and the historical bond transaction data; The similarity sorting unit sorts the similarity between the bond representation vector to be predicted and the historical bond representation vector in the dynamic comparison library, and selects several bond transaction data with the highest similarity to form a similar bond list.

[0016] In a third aspect, the present invention also provides a terminal device comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, wherein the one or more programs include a method for automatically pricing bonds as described in any one of the above.

[0017] In a fourth aspect, the present invention also provides a non-temporary computer-readable storage medium, which, when the instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to execute an automatic bond pricing method as described in any one of the above.

[0018] The present invention discloses a bond automatic pricing method and system, which has the following beneficial effects: Firstly, the present invention does not require complicated steps and realizes automatic pricing of bonds; secondly, the present invention can recommend similar bonds and improve the credibility of pricing results; finally, the present invention can dynamically divide the scope of the bond comparison library, making it more flexible in actual use. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 It is a flow chart of an embodiment of the present invention, showing the various steps of the automatic bond pricing method, including data preprocessing, bond representation calculation, dynamic comparison library construction, similar bond calculation and pricing, etc.

[0021] Figure 2 It is a principle block diagram of an automatic bond pricing system according to an embodiment of the present invention.

[0022] Figure 3 It is a flow chart of an automatic bond pricing system according to an embodiment of the present invention.

[0023] Figure 4 It is a schematic block diagram of an automatic bond pricing system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the description of well-known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present invention.

[0025] Figure 1 This is a flow chart of a bond automatic pricing method according to an embodiment of the present invention, which mainly includes the following steps: Step S1, run the bond representation calculation model and open the bond vector database.

[0026] In the embodiment of the present invention, the training of the bond representation calculation model includes the following steps: preprocessing the historical bond transaction data Sort in ascending order according to transaction time and divide the training set according to a certain ratio and test set In the training set, the bond trading data are those that were traded first, and in the test set, the bond trading data are those that were traded later. The feature extraction network is used to extract features and transform dimensions of bond attributes, and the encoder is used to encode them. Then, the loss function is designed and calculated to perform back-propagation training on the network.

[0027] In the embodiment of the present invention, the bond transaction data and the historical bond representation vector are used to construct a bond vector database. In the specific implementation, the bond representation vector output by the bond representation calculation model is used. , bond trading data , all are sorted in ascending order according to the transaction time, that is, a data structure for quickly finding the bond representation vector corresponding to the bond transaction data is constructed.

[0028] In the embodiment of the present invention, the bond transaction data is the time point data, the i-th data It is the data composed of the attributes of a bond when it is traded at a certain point in time, including the debtor, transaction time, credit rating, yield, bond type, etc. All preprocessing is performed to obtain preprocessed bond trading data .

[0029] Step S2, preprocessing the newly traded bond transaction data to obtain the preprocessed recent newly traded bond transaction data, and updating the bond representation calculation model and bond vector database.

[0030] In an embodiment of the present invention, the newly traded bond transaction data is the bond transaction data on the bond market after the latest transaction time in the bond vector database. The updating of the bond representation calculation model includes using the pre-processed recent newly traded bond transaction data and the original pre-processed bond transaction data in the bond vector database to input the bond representation calculation model for retraining to obtain an updated bond representation calculation model. Updating the bond vector database includes adding the recent newly traded bond transaction data to the bond vector database, and then inputting all the bond transaction data in the bond vector database into the updated bond representation calculation model to obtain an updated bond representation vector, and updating it to the bond vector database.

[0031] In an embodiment of the present invention, the preprocessing of bond data includes data cleaning, converting time attributes into numerical attributes, converting Boolean type attributes into numerical attributes, sequentially encoding category type attributes, normalizing all numerical attributes and sequential encodings, etc.

[0032] Step S2 needs to be performed regularly at regular intervals.

[0033] Step S3, input the preprocessed bond to be predicted into the bond representation calculation model to obtain the bond representation vector to be predicted.

[0034] In the embodiment of the present invention, for the bond data to be predicted First, complete the missing attributes of the bond and preprocess it according to step S2, then input the bond representation calculation model to obtain the bond representation vector to be predicted .

[0035] Step S4, constructing a dynamic comparison library of bonds to be predicted.

[0036] In the embodiment of the present invention, the bond transaction data whose transaction time is before the pricing time of the bond to be predicted and the corresponding bond representation vector are screened out from the bond vector database to construct a dynamic comparison library. For different bonds to be predicted, the dynamic comparison library may be different, which is mainly in line with the real application scenario that only historical bond prices can be used to predict the bond prices traded at a certain point in the future.

[0037] In the embodiment of the present invention, the bond transaction data before the predicted bond transaction time is screened out from the historical bond transaction data, and the corresponding bond representation vector set is extracted. As a dynamic comparison library, for A subset of , which can be quickly determined using a binary search algorithm range.

[0038] Step S5, inputting the representation vector of the bond to be predicted and the dynamic comparison library into the similar bond calculation module, obtaining an ordered list of similar bonds corresponding to the bond to be predicted, and inputting the list into the pricing module to obtain the pricing of the bond to be predicted, including: Set the number of similar bonds N; Calculating the vector similarity between the bond representation vector in the dynamic comparison library and the bond representation vector to be predicted, and sorting them; Select the top N bond transaction data with the highest similarity to form an ordered list of similar bonds of the bond to be predicted; The pricing of the bond to be predicted is obtained based on the ordered list of similar bonds.

[0039] In the embodiment of the present invention, the pricing of the bond to be predicted can be achieved through simple calculation or through prediction by a model.

[0040] The present invention also discloses a bond automatic pricing system, such as Figure 2 As shown, it includes: The bond preprocessing module 100 is used to preprocess the attributes of each bond transaction data, including data cleaning, converting time attributes into numerical attributes, converting Boolean type attributes into numerical attributes, sequentially encoding category type attributes, and normalizing all numerical attributes and sequential encodings; The bond encoding module 200 is used to encode the preprocessed bond trading data, including the preprocessed bond trading data to be predicted and the preprocessed historical bond trading data, and input the preprocessed bond trading data into the bond representation calculation model to obtain the bond representation vector to be predicted and the historical bond representation vector for characterizing the attribute information of the bond; A bond vector database construction module 300 is used to associate the historical bond transaction data with the historical bond representation vector and create an index to construct a bond vector database for fast retrieval of bond transaction data and representation vectors; A comparison library construction module 400 is used to obtain the pricing time of the bond to be predicted, and construct a dynamic comparison library of the bond to be predicted from the bond vector database according to the pricing time; The similar bond calculation module 500 is used to calculate the similarity between the bond representation vector to be predicted and the bond representation vector in the dynamic comparison library, and the similarity is used to characterize the similarity between the bond to be predicted and the historical bond transaction data. The bond transaction data in the dynamic comparison library are sorted based on the similarity, and a number of bond transaction data with the highest similarity are obtained to form an ordered list of similar bonds; A pricing module 600, configured to predict the bond yield to be predicted based on the bond yields in the ordered list of similar bonds; The updating module 700 is used to regularly obtain the newly traded bond transaction data in the recent bond market, pre-process the newly traded bond transaction data to obtain the pre-processed recent newly traded bond transaction data, and update the bond representation calculation model and the bond vector database.

[0041] In an embodiment of the present invention, the bond preprocessing module includes: Data cleaning unit, used to remove duplicate data; Attribute filling unit, used to complete the missing attributes of bonds; A coding unit, used to convert a time attribute into a numerical attribute, convert a Boolean type attribute into a numerical attribute, and perform sequential coding on a category type attribute; Normalization unit, used to normalize all numeric attributes and ordinal encodings.

[0042] In an embodiment of the present invention, the similar coupon calculation module includes: A similarity calculation unit, used to calculate the similarity between the bond representation vector to be predicted and the bond representation vector in the dynamic comparison library, wherein the similarity is used to characterize the similarity between the bond to be predicted and the historical bond transaction data; The similarity sorting unit sorts the similarity between the representation vector of the bond to be predicted and the representation vector in the dynamic comparison library, and selects several bond transaction data with the highest similarity to form a similar bond list.

[0043] In an embodiment of the present invention, a flowchart of a bond automatic pricing system is shown as follows: Figure 3 As shown. First, the historical bond trading data and the bond data to be predicted are input into the bond preprocessing module to obtain the preprocessed historical bond trading data and the preprocessed bond data to be predicted, which are respectively input into the bond encoding module to obtain the historical bond representation vector and the bond representation vector to be predicted. The historical bond trading data, the preprocessed historical bond trading data, and the historical bond representation vector together constitute a bond vector database. According to the pricing time, a dynamic comparison library is constructed from the bond vector database. The bond representation vector to be predicted and the dynamic comparison library are input into the similar bond calculation module to obtain an ordered list of similar bonds, and the ordered list of similar bonds is input into the pricing module to obtain the final pricing of the bond. At the same time, the recent newly traded bond trading data is regularly obtained and input into the update module, and the recent newly traded bond trading data is merged with the existing historical bond trading data, and then used to update the bond representation calculation model and the bond vector database.

[0044] An automatic bond pricing system according to an embodiment of the present invention can be deployed on various platforms or devices, and supports pricing of credit bonds and interest rate bonds within the scope of bond trading data, as well as recommendation of similar bonds. A schematic block diagram of an automatic bond pricing system according to an embodiment of the present invention is shown in FIG. Figure 4 shown.

[0045] An embodiment of the present invention discloses a computer-readable storage medium, which includes a computer program. When the computer program runs on an electronic device, the electronic device executes an automatic bond pricing method as described in any one of the above items.

[0046] An embodiment of the present invention discloses a device, which includes a connected processor and a memory, wherein the processor is used to execute a computer program stored in the memory to perform an automatic bond pricing method as described in any one of the above items.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A bond automatic pricing method, characterized in that: It includes the following steps: Step S1, running the bond representation calculation model and opening the bond vector database; Step S2, preprocessing the newly traded bond transaction data to obtain the preprocessed recent newly traded bond transaction data, and updating the bond representation calculation model and bond vector database; Step S3, inputting the preprocessed bond to be predicted into the bond representation calculation model to obtain a bond representation vector to be predicted; Step S4, constructing a dynamic comparison library of bonds to be predicted; Step S5, input the representation vector of the bond to be predicted and the dynamic comparison library into the similar bond calculation module, obtain an ordered list of similar bonds corresponding to the bond to be predicted, input it into the pricing module, and obtain the pricing of the bond to be predicted.

2. The automatic bond pricing method according to claim 1, characterized in that: In step S1, the training process of the bond representation calculation model includes: Acquire historical bond trading data from the bond market, and preprocess the historical bond trading data to obtain preprocessed historical bond trading data; The pre-processed historical bond trading data is input into a deep learning neural network for training, and a bond representation calculation model is obtained after the training is completed.

3. The automatic bond pricing method according to claim 2, characterized in that: In step S1, the process of constructing the bond vector database includes: Inputting the preprocessed historical bond transaction data into the bond representation calculation model to obtain a historical bond representation vector; The historical bond transaction data and the historical bond representation vectors are associated and indexed to construct a bond vector database.

4. The automatic bond pricing method according to claim 3, characterized in that: The historical bond trading data is time point data, and each piece of data is the attribute of a certain bond when it is traded at a time point.

5. The automatic bond pricing method according to claim 1, characterized in that: In step S2, the preprocessing step includes data cleaning, converting time attributes into numerical attributes, converting Boolean type attributes into numerical attributes, sequentially encoding category type attributes, and normalizing all numerical attributes and sequential encodings; Step S2 needs to be performed regularly at regular intervals.

6. The automatic bond pricing method according to claim 1, characterized in that: In step S4, constructing the dynamic comparison library of the bonds to be predicted includes screening out bond transaction data whose transaction time is before the pricing time of the bonds to be predicted and their corresponding bond representation vectors from the bond vector database to construct the dynamic comparison library.

7. The automatic bond pricing method according to claim 1, characterized in that: In step S5, the bond representation vector to be predicted and the dynamic comparison library are input into the similar bond calculation module to obtain an ordered list of similar bonds corresponding to the bond to be predicted, including: Set the number of similar bonds N; Calculating the vector similarity between the bond representation vector in the dynamic comparison library and the bond representation vector to be predicted, and sorting them; The top N bond transaction data with the highest similarity are selected to form an ordered list of similar bonds of the bond to be predicted.

8. A bond automatic pricing system, characterized in that: include: The bond preprocessing module is used to preprocess the attributes of each bond transaction data, including data cleaning, converting time attributes into numerical attributes, converting Boolean type attributes into numerical attributes, sequentially encoding category type attributes, and normalizing all numerical attributes and sequential encodings; A bond encoding module is used to encode the preprocessed bond trading data, including the preprocessed bond trading data to be predicted and the preprocessed historical bond trading data, and input the preprocessed bond trading data into the bond representation calculation model to obtain the bond representation vector to be predicted and the historical bond representation vector for characterizing the attribute information of the bond; A bond vector database construction module associates the historical bond transaction data with the historical bond representation vector and creates an index to construct a bond vector database for fast retrieval of bond transaction data and representation vectors; A comparison library construction module, used to obtain the pricing time of the bond to be predicted, and construct a dynamic comparison library of the bond to be predicted from the bond vector database according to the pricing time; A similar bond calculation module is used to calculate the similarity between the bond representation vector to be predicted and the bond representation vector in the dynamic comparison library, and the similarity is used to characterize the similarity between the bond to be predicted and the historical bond transaction data; based on the similarity, the bond transaction data in the dynamic comparison library are sorted to obtain a number of bond transaction data with the highest similarity to form an ordered list of similar bonds; A pricing module, used for predicting the bond yield to be predicted based on the bond yields in the ordered list of similar bonds; The update module is used to regularly obtain the newly traded bond transaction data in the recent bond market, pre-process the newly traded bond transaction data to obtain the pre-processed recent newly traded bond transaction data, and update the bond representation calculation model and the bond vector database.

9. A terminal device, characterized in that: The terminal device includes a memory, a processor, and a program of an automatic bond pricing method stored in the memory and executable on the processor. When the processor executes the program of the automatic bond pricing method, the steps of the automatic bond pricing method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program for an automatic bond pricing method. When the program for an automatic bond pricing method is executed by a processor, the steps of an automatic bond pricing method as described in any one of claims 1 to 7 are implemented.

Images