Coupon source scarcity evaluation method and device

Abstract

The invention discloses a coupon source scarcity evaluation method and device. The method comprises the following steps: acquiring market information source data; off-line factors are calculated, and a daily frequency factor library is generated; real-time factors are generated and aggregated through a self-developed streaming computing framework, and an intra-day factor library is generated; modeling by adopting an integrated decision tree model, and generating an off-line prediction model and an on-line prediction model through model training by taking the proportion of the sales volume of the fused coupons to the circulation market value of the fused coupons as a scarcity characterization index; and fusing the generated offline prediction model and the online prediction model by using a model fusion method to generate a coupon source scarcity degree evaluation result. According to the method, factors related to the scarcity degree of the fusion coupons are screened out by taking the proportion of the sale amount of the fusion coupons to the circulation market value of the fusion coupons as a key index; the integrated decision tree model is utilized to construct the offline and online prediction models, the model fusion method is utilized to integrate the advantages of the two models, an accurate coupon source scarcity degree evaluation result is generated, and the problem that the evaluation of the coupon source scarcity degree is incomplete and inaccurate in the prior art is comprehensively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning and financial data analysis, and in particular relates to a method and device for evaluating the scarcity of bond sources. Background Art

[0002] In the securities market, accurately assessing the scarcity of securities sources plays a crucial role in brokerage firms' decision-making regarding securities pool operations. However, existing assessment methods have significant shortcomings in acquiring and processing market data on securities lending targets and using this data to assess the scarcity of securities sources.

[0003] On the one hand, while subjective evaluations by salespeople can incorporate their experience and understanding of the market, this approach relies too heavily on personal judgment and lacks objectivity and systematicity. Furthermore, due to the limited metrics available to salespeople, it's difficult to fully encompass the complex factors influencing bond scarcity, resulting in potentially inaccurate and incomplete evaluation results. On the other hand, while objective evaluation methods based on historical market data do account for the continuity of bond scarcity, they rely primarily on historical data and lack sensitivity to real-time market fluctuations. While this approach may be valuable in relatively stable market conditions, it can be slow to react to sudden market fluctuations and fail to reflect the latest market developments, leading to delayed and inaccurate evaluation results.

[0004] More critically, existing technologies cannot effectively acquire and process daily and real-time market data for securities lending targets. Daily market data provides an overview of the underlying security's performance over a period of time, while real-time market data reflects real-time market dynamics. Without this data, existing evaluation methods struggle to comprehensively and accurately assess security scarcity. Summary of the Invention

[0005] The technical problem to be solved by the present invention is: how to effectively obtain and process the daily market source data and real-time market source data of the securities lending target, and use these data to accurately evaluate the scarcity of the securities source.

[0006] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solutions.

[0007] The present invention provides a method for evaluating the scarcity of a voucher source, comprising:

[0008] Obtain daily and real-time market data for margin lending targets;

[0009] Utilize the daily market source data and combine it with the high-performance offline computing capabilities of the big data platform to calculate offline factors;

[0010] Generate a daily frequency factor library required for an offline prediction model based on the offline factors;

[0011] Based on the real-time market source data, real-time factors are generated and aggregated through a self-developed streaming computing framework;

[0012] Based on the real-time factors, the intraday factor library required for the online prediction model is calculated and generated in real time;

[0013] An integrated decision tree model is used for modeling. The ratio of margin sales to the circulating market value of margin loans is used as a scarcity indicator. Daily factors highly correlated with margin scarcity are identified in the daily factor library and used as feature inputs for the integrated decision tree model. An offline prediction model is generated through model training. Intraday factors highly correlated with margin scarcity are identified in the intraday factor library and used as feature inputs for the integrated decision tree model. An online prediction model is generated through model training.

[0014] Using the model fusion method, the generated offline prediction model and online prediction model are integrated to generate a daily coupon scarcity evaluation result report.

[0015] Based on the aforementioned security source scarcity evaluation method, the short selling scarcity characterization indicator is the ratio of short selling sales to the short selling circulating market value.

[0016] Based on the aforementioned coupon scarcity evaluation method, the offline ETL algorithm is used to calculate the offline factor, and the real-time factor is calculated using the real-time online algorithm.

[0017] Based on the aforementioned security source scarcity evaluation method, the modeling process of the online prediction model also includes aggregating and reducing the frequency of the real-time market source data of the securities lending target, and processing the generated real-time factors.

[0018] Based on the aforementioned security source scarcity evaluation method, the offline prediction model initializes the score of security source scarcity at the beginning of the day based on offline factors; the online prediction model captures individual stocks with large market fluctuations based on the real-time market conditions of the day, and promptly adjusts and optimizes the scarcity score given by the offline prediction model.

[0019] Based on the aforementioned security scarcity evaluation method, by adjusting the time step of the aggregated real-time factors, intraday market monitoring indicators are generated to capture individual stocks with large market fluctuations. The market monitoring indicators include dynamic volatility, capital flow indicators, trading activity indicators and price breakthrough indicators.

[0020] Based on the aforementioned security source scarcity evaluation method, during the model training process, the rankIC value of the securities lending scarcity index is used to evaluate the effectiveness of the factors in the daily frequency factor library and the intraday factor library, and to identify factors that are highly correlated with securities lending scarcity.

[0021] Based on the aforementioned bond scarcity evaluation method, the generated offline prediction model and online prediction model are fused using the Stacking model fusion method to generate the bond scarcity evaluation result.

[0022] Based on the aforementioned coupon source scarcity evaluation method, the offline prediction model and the online prediction model are fused to generate a coupon source evaluation learner, and the coupon source evaluation learner is deployed in a production environment.

[0023] The present invention also provides a device for evaluating the scarcity of a voucher source, comprising:

[0024] The data acquisition module is used to obtain daily and real-time market data of securities lending targets;

[0025] An offline factor calculation module is used to calculate offline factors by utilizing the daily market source data and combining the offline computing capabilities of the big data platform;

[0026] A daily frequency factor library generation module, which is used for generating the daily frequency factor library required by the offline factor generation offline prediction model;

[0027] A real-time factor calculation module, configured to generate and aggregate real-time factors based on the real-time market source data through a streaming computing framework;

[0028] An intraday factor library generation module is used to calculate and generate the intraday factor library required by the online prediction model in real time based on the real-time factors;

[0029] The modeling and training module is used to build models using an integrated decision tree model. The module uses the ratio of margin sales to the market value of margin trading as a scarcity indicator. Daily factors highly correlated with margin trading scarcity are identified from the daily factor library and used as feature inputs into the integrated decision tree model. This training generates an offline prediction model. Intraday factors highly correlated with margin trading scarcity are identified from the intraday factor library and used as feature inputs into the integrated decision tree model. This training generates an online prediction model.

[0030] The model fusion module is used to use the model fusion method to fuse the generated offline prediction model and the online prediction model to generate a daily coupon scarcity evaluation result report.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] The present invention analyzes and mines the daily frequency factors and intraday factors of the stock market, and uses the ratio of the amount of securities sold to the circulating market value of securities as a scarcity characterization index to screen out daily frequency factors and intraday factors that are highly correlated with the scarcity of securities. Utilizing the screened daily frequency factors and intraday factors and an integrated decision tree model, an offline prediction model and an online prediction model are established to explore the intrinsic connection between security scarcity and stock market factors. The offline prediction model and the online prediction model can provide highly reliable predictions based on real-time market information, allowing investors and securities lending business operators to conduct retrospective analysis based on historical data, as well as to make forward-looking predictions based on real-time data, thereby flexibly adjusting strategies. In addition, the generated offline prediction model and online prediction model are fused using a model fusion method, and the generated daily security scarcity evaluation results can accurately evaluate the security scarcity at the intraday and hourly levels. This solves the technical problem of the existing technology that cannot comprehensively and accurately evaluate security scarcity. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is an overall framework diagram of a device for evaluating scarcity of a voucher source according to an embodiment of the present invention;

[0034] Figure 2 This is a modeling flow chart of an online prediction model of a method for evaluating the scarcity of a coupon source according to an embodiment of the present invention;

[0035] Figure 3 This is a flow chart of a model fusion module of a device for evaluating scarcity of a voucher source according to an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram of the Stacking model fusion method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] Example 1

[0038] This embodiment introduces a method for evaluating the scarcity of a voucher source, including:

[0039] Obtain daily market data for margin trading targets, including opening price, highest price, lowest price, closing price, change, trading volume, turnover, price-to-earnings ratio, candlestick chart, and moving average;

[0040] Import daily market source data into the big data platform, clean and pre-process it, remove outliers, missing values, and duplicate values, and perform standardization operations to ensure data accuracy and completeness. Leverage the high-performance offline computing capabilities of the big data platform to perform statistical analysis and feature extraction on daily market source data, and use offline ETL algorithms to calculate offline factors.

[0041] Generate the daily frequency factor library required for the offline forecasting model based on the calculated offline factors.

[0042] Obtain real-time market data for margin lending targets, including opening price, highest price, lowest price, closing price, price change, trading volume, turnover, price-to-earnings ratio, real-time candlestick chart, MACD indicator, and RSI indicator. Use an efficient aggregation and frequency reduction calculation strategy to obtain real-time market data for margin lending targets to ensure real-time calculation performance.

[0043] Based on real-time market source data, real-time factors are generated and aggregated using real-time online computing through a streaming computing framework.

[0044] Process the generated and aggregated real-time factors, and calculate and generate the intraday factor library required for the online prediction model in real time.

[0045] The Bagging ensemble decision tree model is used for modeling, and the proportion of securities lending sales to the securities lending market value is used as the scarcity representation indicator. The rankIC value of the securities lending scarcity indicator is used to evaluate the effectiveness of the factors in the daily frequency factor library and the intraday factor library. The daily frequency factors and intraday factors that are highly correlated with the scarcity of securities lending are identified and used as features to input the Bagging ensemble decision tree model respectively. Cross-validation technology is used to evaluate the performance of the model during the model training process, and the optimal model parameters are selected to generate the offline prediction model and the online prediction model respectively.

[0046] The offline prediction model is based on offline factors and initializes the scoring of the scarcity of securities at the beginning of the day; the online prediction model is based on the real-time market conditions of the day, adjusts the time step of the aggregated real-time factors, and generates intraday market monitoring indicators including dynamic volatility, capital flow indicators, trading activity indicators and price breakthrough indicators.

[0047] Reasonable threshold settings and weight distribution are assigned to market monitoring indicators to capture individual stocks with significant market fluctuations, and the scarcity scores given by the offline prediction model are adjusted and optimized in a timely manner. The market monitoring indicators update the features and parameters of the models in a timely manner according to the prediction results of the offline prediction model and the online prediction model and the actual market conditions, thereby improving the accuracy and stability of the two models.

[0048] Using Stacking model fusion method, such as Figure 3 、 Figure 4 As shown, the generated offline prediction model and online prediction model are integrated to generate a coupon source evaluation learner and deployed in the production environment to generate a daily coupon source scarcity evaluation result report. The generated daily coupon source scarcity evaluation result report can accurately evaluate the coupon source scarcity at the daily and hourly levels.

[0049] When analyzing the refinancing and securities lending markets, daily securities scarcity evaluation reports are used to assist in market analysis and decision-making.

[0050] When measuring and analyzing the profit and loss of the bond pool, and forecasting it, we will take stock of the scarcity distribution of the bond pool based on the daily bond scarcity evaluation report, measure and predict the profit and loss of the bond pool, and assist in business decision-making.

[0051] In the pricing of bond sources, daily bond scarcity evaluation results reports are used to assist bond source pricing.

[0052] Example 2

[0053] Based on the same inventive concept as Example 1, this example introduces a device for evaluating the scarcity of a coupon source. Figure 1 Shown, including:

[0054] The data acquisition module is used to obtain daily and real-time market data of securities lending targets;

[0055] An offline factor calculation module is used to calculate offline factors by utilizing the daily market source data and combining the offline computing capabilities of the big data platform;

[0056] A daily frequency factor library generation module, which is used for generating the daily frequency factor library required by the offline factor generation offline prediction model;

[0057] A real-time factor calculation module, configured to generate and aggregate real-time factors based on the real-time market source data through a streaming computing framework;

[0058] An intraday factor library generation module is used to calculate and generate the intraday factor library required by the online prediction model in real time based on the real-time factors;

[0059] The modeling and training module is used to build models using an integrated decision tree model. The module uses the ratio of margin sales to the market value of margin trading as a scarcity indicator. Daily factors highly correlated with margin trading scarcity are identified from the daily factor library and used as feature inputs into the integrated decision tree model. This training generates an offline prediction model. Intraday factors highly correlated with margin trading scarcity are identified from the intraday factor library and used as feature inputs into the integrated decision tree model. This training generates an online prediction model.

[0060] The model fusion module is used to use the model fusion method to fuse the generated offline prediction model and the online prediction model to generate a daily coupon scarcity evaluation result report.

[0061] The specific functional implementation of each of the above modules can be found in the relevant content of the method in Example 1 and will not be elaborated on here.

[0062] In summary, the present invention analyzes and mines the daily frequency factors and intraday factors of the stock market, and uses the ratio of the amount of securities sold to the circulating market value of securities as a scarcity characterization index to screen out daily frequency factors and intraday factors that are highly correlated with the scarcity of securities. Utilizing the screened daily frequency factors and intraday factors and an integrated decision tree model, an offline prediction model and an online prediction model are established to explore the intrinsic connection between security scarcity and stock market factors. The offline prediction model and the online prediction model can provide highly reliable predictions based on real-time market information, allowing investors and securities lending business operators to conduct retrospective analysis based on historical data and to make forward-looking predictions based on real-time data, thereby flexibly adjusting strategies. In addition, using the model fusion method, the generated offline prediction model and online prediction model are fused, and the generated daily security scarcity evaluation results can accurately evaluate the security scarcity at the intraday and hourly levels. This solves the technical problem of the existing technology that cannot comprehensively and accurately evaluate security scarcity.

[0063] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0064] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which are all protected by the present invention.

Claims

1. A method for evaluating the scarcity of a voucher source, characterized in that: include: Obtain daily and real-time market data for margin lending targets; Utilizing the daily market data and combining it with the offline computing capabilities of the big data platform, the offline factors are calculated; Generate a daily frequency factor library required for an offline prediction model based on the offline factors; Based on the real-time market source data, real-time factors are generated and aggregated through a streaming computing framework; Based on the real-time factors, the intraday factor library required for the online prediction model is calculated and generated in real time; An integrated decision tree model is used for modeling. The ratio of margin sales to the circulating market value of margin loans is used as a scarcity indicator. Daily factors highly correlated with margin scarcity are identified in the daily factor library and used as feature inputs for the integrated decision tree model. An offline prediction model is generated through model training. Intraday factors highly correlated with margin scarcity are identified in the intraday factor library and used as feature inputs for the integrated decision tree model. An online prediction model is generated through model training. Using the model fusion method, the generated offline prediction model and online prediction model are integrated to generate a daily coupon scarcity evaluation result report.

2. The method for evaluating the scarcity of a voucher source according to claim 1, characterized in that: The indicator representing the scarcity of securities lending is the ratio of the amount of securities lending sold to the circulating market value of securities lending.

3. The method for evaluating the scarcity of a voucher source according to claim 1, characterized in that: Offline factors are calculated using offline ETL algorithms, and real-time factors are calculated using real-time online algorithms.

4. The method for evaluating the scarcity of a voucher source according to claim 1, characterized in that: The modeling process of the online prediction model also includes aggregating and reducing the frequency of the real-time market source data of the securities lending target and processing the generated real-time factors.

5. The method for evaluating the scarcity of a voucher source according to claim 1, characterized in that: The offline prediction model initializes the scarcity score of the security source at the beginning of the day based on offline factors; the online prediction model captures individual stocks with large market fluctuations based on the real-time market conditions of the day, and promptly adjusts and optimizes the scarcity score given by the offline prediction model.

6. The method for evaluating the scarcity of a voucher source according to claim 5, characterized in that: By adjusting the time step of the aggregated real-time factors, intraday market monitoring indicators are generated to capture individual stocks with large market fluctuations. The market monitoring indicators include dynamic volatility, capital flow indicators, trading activity indicators and price breakthrough indicators.

7. The method for evaluating the scarcity of a voucher source according to claim 1, characterized in that: During the model training process, the rankIC value of the securities lending scarcity index is used to evaluate the effectiveness of the factors in the daily factor library and the intraday factor library, and to identify factors that are highly correlated with securities lending scarcity.

8. The method according to claim 1, characterized in that The generated offline prediction model and online prediction model are fused using the Stacking model fusion method to generate a bond scarcity evaluation result that includes bond scarcity evaluation, prediction trend and market factor analysis.

9. The method for evaluating the scarcity of a voucher source according to claim 8, characterized in that: The offline prediction model and the online prediction model are integrated to generate a coupon source evaluation learner, and the coupon source evaluation learner is deployed in a production environment.

10. A device for evaluating the scarcity of a voucher source, characterized in that: include: The data acquisition module is used to obtain daily and real-time market data of securities lending targets; An offline factor calculation module is used to calculate offline factors by utilizing the daily market source data and combining the offline computing capabilities of the big data platform; A daily frequency factor library generation module, which is used for generating the daily frequency factor library required by the offline factor generation offline prediction model; A real-time factor calculation module, configured to generate and aggregate real-time factors based on the real-time market source data through a streaming computing framework; An intraday factor library generation module is used to calculate and generate the intraday factor library required by the online prediction model in real time based on the real-time factors; The modeling and training module is used to build models using an integrated decision tree model. The module uses the ratio of margin sales to the market value of margin trading as a scarcity indicator. Daily factors highly correlated with margin trading scarcity are identified from the daily factor library and used as feature inputs into the integrated decision tree model. This training generates an offline prediction model. Intraday factors highly correlated with margin trading scarcity are identified from the intraday factor library and used as feature inputs into the integrated decision tree model. This training generates an online prediction model. The model fusion module is used to use the model fusion method to fuse the generated offline prediction model and the online prediction model to generate a daily coupon scarcity evaluation result report.

Images