Transaction settlement prediction system and method

Abstract

A computer-implemented method for transaction settlement prediction may include receiving data for a plurality of past financial transactions, training a machine learning model using the data for the plurality of past financial transactions, receiving one or more parameters of a subject financial transaction of the plurality of recently performed financial transactions, the trained machine learning model is utilized to determine the possibility that the subject financial transaction will fail, the trained machine learning model is utilized to determine the most probable reason that the subject financial transaction will fail, and the possibility that the subject financial transaction will fail and the most probable reason that the subject financial transaction will fail are presented to a user.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of priority to U.S. Application No. 17 / 935,506, filed on September 26, 2022, which claims the benefit of priority to Indian Application No. 202211006618, filed on February 8, 2022, which are incorporated herein by reference in their entirety and become a part of the present invention. Technical Field

[0003] Various embodiments of the present disclosure relate generally to processing financial transactions and, more particularly, to transaction settlement failure prediction. Background Art

[0004] Capital markets firms are experiencing an increasing volume of complex, high-risk transactions, such as in equity and derivatives markets. Some of these transactions may result in settlement failures due to a variety of factors and may require intensive monitoring and tracking to prevent settlement failures. These potential issues may require support team agents to continuously monitor and track transactions, which is time-consuming and costly for capital markets firms. For example, non-automated monitoring of a large number of pending transactions without in-depth understanding of which transactions are most likely to result in settlement failures may result in resources being expended on transactions that fail to settle after execution, or may result in resources being expended on monitoring transactions that are not in danger of settlement failures. These resources may include both manual labor and technical resources (computing time, memory, and other storage) used in the settlement monitoring and mitigation process.

[0005] The present disclosure is directed to overcoming one or more of the challenges set forth above. Summary of the invention

[0006] According to certain aspects of the present disclosure, systems and methods for transaction settlement prediction are disclosed.

[0007] In one embodiment, a computer-implemented method for transaction settlement prediction is disclosed, the method comprising: receiving data of multiple past financial transactions, training a machine learning model using the data of the multiple past financial transactions, receiving one or more parameters of a subject financial transaction in multiple recently executed financial transactions, determining a probability that the subject financial transaction will fail using the trained machine learning model, determining a most likely reason why the subject financial transaction will fail using the trained machine learning model, and presenting to a user the probability that the subject financial transaction will fail and the most likely reason why the subject financial transaction will fail.

[0008] According to another embodiment, a system for transaction settlement prediction is disclosed, the system comprising: a data storage device storing instructions for transaction settlement prediction in an electronic storage medium; a processor executing the instructions to perform a method comprising the following steps: receiving data of multiple past financial transactions, training a machine learning model using the data of the multiple past financial transactions, receiving one or more parameters of a subject financial transaction in multiple recently executed financial transactions, determining a probability that the subject financial transaction will fail using the trained machine learning model, determining the most likely reason why the subject financial transaction will fail using the trained machine learning model, and presenting the probability that the subject financial transaction will fail and the most likely reason why the subject financial transaction will fail to a user.

[0009] According to another embodiment, a non-transitory machine-readable medium storing instructions, when the instructions are executed by a computing system, can cause the computing system to execute a transaction settlement prediction method, the method comprising: receiving data of multiple past financial transactions, training a machine learning model using the data of the multiple past financial transactions, receiving one or more parameters of a subject financial transaction in multiple recently executed financial transactions, determining a probability that the subject financial transaction will fail using the trained machine learning model, determining a most likely reason why the subject financial transaction will fail using the trained machine learning model, and presenting to a user the probability that the subject financial transaction will fail and the most likely reason why the subject financial transaction will fail.

[0010] Further objects and advantages of the disclosed embodiments will be partially set forth in the following description and partially apparent, or may be learned through the practice of the disclosed embodiments. These objects and advantages of the disclosed embodiments may be realized and obtained by utilizing the elements and combinations particularly pointed out in the appended claims.

[0011] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosed embodiments, as claimed. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings, which are included in this specification and constitute a part of this specification, illustrate various exemplary embodiments and, together with the following description, explain the principles of the disclosed embodiments.

[0013] Figure 1A A machine learning framework for transaction settlement prediction according to one or more embodiments is described.

[0014] Figure 1B An exemplary system infrastructure for transaction settlement prediction is depicted in accordance with one or more embodiments.

[0015] Figure 2Data related to failed transaction settlement in a system for transaction settlement prediction according to one or more embodiments is depicted.

[0016] Figure 3 A data model in a system for transaction settlement prediction according to one or more embodiments is depicted.

[0017] Figure 4 A process flow for training a machine learning model in a system for transaction settlement prediction according to one or more embodiments is described.

[0018] Figure 5 Depicted is performance of a machine learning model in a system for transaction settlement prediction according to one or more embodiments.

[0019] Figure 6 A flow chart of a method for transaction settlement prediction according to one or more embodiments is depicted.

[0020] Figure 7-9 A user interface for presenting risky transaction settlement in a system for transaction settlement prediction according to one or more embodiments is described. DETAILED DESCRIPTION

[0021] Various embodiments of the present disclosure are generally directed to enabling voice control of an interactive audio-visual environment, and monitoring user behavior to assess engagement.

[0022] Although the terms used below are used in conjunction with certain specific examples of the present disclosure, these terms are to be interpreted in the broadest sense. In fact, certain terms may even be emphasized below; however, this detailed description section will publicly and specifically define any term that is intended to be interpreted in any limited manner.

[0023] As described above, capital markets firms are experiencing an increasing volume of complex, high-risk transactions, such as in equity and derivatives markets. Due to a variety of factors, some of these transactions may result in settlement failures and may require in-depth monitoring and tracking to prevent settlement failures. These potential issues may require support team agents to continuously monitor and track transactions, which is time-consuming and costly for capital markets firms. For example, non-automated monitoring of a large number of pending transactions without in-depth understanding of which transactions are most likely to result in settlement failures may result in resources being spent on transactions that fail to settle after execution, or may result in resources being spent on monitoring transactions that are not in danger of settlement failures. As described below, one or more embodiments are directed to overcoming one or more of these challenges. In particular, one or more embodiments may generate an estimated probability that a particular transaction may fail to settle, as well as the most likely cause of such failure. For example, by providing a prediction that a financial transaction is highly likely to fail to settle, one or more embodiments may alert a broker / dealer to prompt the broker / dealer to take action, which may reduce the broker / dealer's financial risk and detect risky transactions earlier. For example, by providing predictions of the most likely reasons why a financial transaction may fail to settle, potential failures can be mitigated more quickly, clues can be provided for investigating potentially risky transactions, the cyclical impact of the transaction settlement process can be reduced, and transaction failures can be analyzed. These benefits can further provide brokers / dealers with reduced transaction processing costs, savings in automation and labor costs, reduced borrowing costs, improved reputation, and improved funding estimates. Resources saved through such predictions can include manual labor and technical resources (computing time, memory, and other storage) used in the settlement monitoring and mitigation process. For transactions that have a low probability of settlement failure and therefore do not require monitoring or mitigation, the resources consumed can be reduced.

[0024] One or more embodiments may utilize machine learning models to predict the probability that a settlement failure may occur for a particular transaction, as well as the most likely reason for the settlement failure. Figure 1A A machine learning framework for transaction settlement prediction according to one or more embodiments is described. Figure 1AAs shown, the machine learning module 110 can interact with other modules in the transaction settlement prediction framework. For example, such a framework may include: an open application programming interface (API) 120, through which transaction details and prediction data can be programmatically accessed; a database 130, which can access transactions, predictions, and historical records and can be used as an information source for applications; a failure monitoring module 140, which can extend functionality by providing failure prediction services; a workflow automation module 150, which can automatically assign key predictions to users for monitoring, tracking, and other operations, and perform workflow management of transaction monitoring tasks; a securities financing module 160, which can use settlement predictions and other data (such as data from database 130) to predict borrowing needs ( The clearing and settlement module 180 may support the modernization of settlement failure management functions and may leverage valuable data, such as using settlement forecasts and other data (such as data from the database 130) to provide new products to new or existing clients and customers, such as intelligent monitoring of pending trade settlements, and further automation of existing settlement review and mitigation workflows. These products may provide dealer / broker clients with intelligent decision making integrated into existing manual review workflows to speed up throughput and minimize risk and / or exposure. In addition, the business intelligence module 190 may generate a user interface (UI) to present forecasts and related risk indicators, as well as generate settlement risk event notifications, such as violations of risk level thresholds or limits. The details of the operation of these components will be further described below.

[0025] The machine learning model may be a general model for all financial transactions, or a specific model for a transaction customer, or a further specialized model for a transaction type or any combination of transaction attributes.

[0026] Figure 1A The machine learning framework for transaction settlement prediction can be embodied as a computer system, such as Figure 1B The system shown. Figure 1B As shown, the system may include a settlement predictor 102. The settlement predictor 102 may include a settlement prediction engine 104, which may include modules for generating transaction settlement predictions, including but not limited to a machine learning module 110. As described above, the settlement predictor 102 may further include an open API 120, a failure monitoring module 140, a workflow automation module 150, a securities finance module 160, an analysis module 170, a clearing and settlement module 180, and a business intelligence module 190.

[0027] Database 130 may include data related to currently executed transactions (TD) in database 132, currently settled day (SD) transactions in database 134, transactions one day after the settlement date (SD-1) in database 136, and transactions two days after the settlement date (SD-2) in database 138. Database 130 may further include data related to past executed transactions. As described below, the data for each transaction represented in database 130 may be parameters of each transaction, as well as a calculated likelihood of settlement failure for the transaction, the most likely reason for settlement failure, the actual failure or settlement date of the transaction, and the reason for the failure of the transaction.

[0028] The business intelligence module 190 may present the user interface 106 to the transaction user 122. The user interface 106 may provide the user (e.g., the transaction user 122) with the probability of settlement failure of the pending financial transaction and / or detailed information about the probability of settlement failure of the pending financial transaction. Figure 7-9 Explore the details of an exemplary user interface.

[0029] The analysis module 170 may generate reports for viewing by the trading user 122. For example, the report may include a summary and trend of trade settlement failures, as well as detailed information on specific trades that have failed in the past or have reached a failure risk probability threshold. The report may be customized based on the specific needs of a user (e.g., a trading user 122 or a broker-dealer). In addition, the report may be a static report, or may be a dynamic report presented in an interactive electronic format, possibly including dynamic visualizations to provide an intuitive view of pending trade settlements and the risk of failed settlements.

[0030] Transaction users 122 may access API 120 to directly access data, including data from database 130 or data generated by settlement prediction engine 104 , or access control and user-specified settings of settlement prediction engine 102 .

[0031] User-specified settings for the settlement predictor 102 (e.g., settings that may be provided via the API 120, a configuration file, or via other user interfaces) may include refresh frequencies for trade evaluations, retraining frequencies for models, settings for selecting a time window of past trades for training the model, or settings for tuning model performance by pruning the trades selected for training the model or the trades monitored by the trained model. Pruning may be performed based on trade type, trade quantity or net amount, security type or sector, client, client group, or client sector, etc. In addition, the machine learning model may be tuned by setting a discrimination threshold below which the model determines that a trade may not settle.

[0032] The machine learning model 110 may be a binary classifier, for example, generated using an XGBoost algorithm with a binary:logistic target. For example, tuning parameters of the binary classifier generator may include the number of estimators, the learning rate, the subsample ratio of training instances, the maximum depth, the subsample ratio of columns, the minimum subelement weight, the regularization term of the weight, the balance of positive and negative weights, etc. The tuning parameters may be provided through the API 120, a configuration file, or through other user interfaces.

[0033] In the context of Machine Learning Module 110, we discuss Figure 1A and 1B However, settlement failure risk analysis need not be limited to machine learning algorithms. Other analysis algorithms may also be used, such as statistical models including regression analysis and artificial intelligence including neural networks.

[0034] The settlement prediction system 100 may be used to analyze a flow of financial transactions, some of which may fail to settle due to one or more reasons. Figure 2 Figure 2 depicts data 200 related to the flow of financial transactions over a three-month period. Figure 2 As shown, for each day 230 , there may be a certain number of executed transactions 220 and a certain number of settlement failed transactions 210 .

[0035] Settlement predictor 102 may receive data related to financial transaction flow 200 via database 130 and may generate a data model representing the attributes of each transaction and the intelligence generated by settlement predictor 102 . Figure 3 Such a data model 300 is depicted. Figure 3 As shown, the data model 300 may include one or more data sets 310, each of which may include data of current and / or past tracked financial transactions, which may include input data 320 (which may include attributes of each tracked financial transaction) and smart data 330 (which may include attributes generated by the settlement prediction engine 104, such as the probability of settlement failure, settlement prediction, and the most likely cause of the predicted settlement failure, etc.). The input data 320 may include static attributes 340, dynamic attributes 350, and account and security attributes 360. The data model 300 may include a separate record 370 for each tracked financial transaction.

[0036] Static attributes 340 do not change during the life cycle of a transaction. Static attributes 340 may include a customer account identifier, a transaction security identifier, a transaction quantity or amount, a transaction price, a transaction execution and / or settlement date, transaction proceeds delivery instructions, an account type, whether the transaction is a buy transaction or a sell transaction, a counterparty, a security type or subtype, etc.

[0037] Dynamic attributes 350 may change during the life cycle of a transaction, and may include the status of the transaction matching with the counterparty, transaction confirmation, cancellation or correction, transaction status, transaction allocation, Depository Trust Company (DTC) status code, current market price of the transaction, transaction risk exposure, foreign exchange rate conversion, etc.

[0038] Account and security attributes 360 may also change during the life cycle of a transaction and may include the trading account's cash and money market balances, stock records, account liquidation value, the trading account's recent failure history, the counterparty's recent failure history, any outstanding calls on the trading account or counterparty's accounts, the ease with which securities can be lent or borrowed, etc.

[0039] The settlement predictor 102 may use the input data 320 of the data model 300 to train the machine learning module 110 of the settlement prediction engine 104 . Figure 4 A process flow for training a machine learning model 110, such as a process flow that may be performed by a settlement predictor 102, is depicted according to one or more embodiments. Figure 4 As shown, the data model 300 may be divided into a training data set 410, which may include a first portion of data from the data model 300, such as one-third of the data from the data model 300 related to past transactions (including transactions that have been settled or failed to settle). At least a portion of the remaining data from the data model 300 related to past transactions may be allocated to the test data set 460. The portion of the remaining data from the data model 300 allocated to the test data set 460 may be the entire remaining portion, or may be less than the entire remaining portion. For example, the test data set 460 may exclude transactions within a predetermined time period after the last transaction executed in the training data set 410, such as a week, a month, or some other time period. The time period may be automatically determined, or the time period may be set by a configuration specified by a user. In operation 420, the settlement predictor 102 may normalize the data in the training data set 410 and may adjust one or more machine learning algorithm parameters, such as a discrimination threshold at which the model determines that a transaction may not be settled. In operation 430, the settlement predictor 102 may construct a plurality of machine learning models based on the normalized training data set 410 and one or more machine learning algorithm parameters. That is, the training data 410 may be used to construct N machine learning models 455, namely, model M 1 -M nIn operation 440, the settlement predictor 102 may review the results of each model 455 and may readjust one or more machine learning algorithm parameters. Operation 440 may be a fully automated process, or may include some human intervention to evaluate the results of the constructed models and manually adjust the algorithm parameters. Operations 420-440 may be iteratively performed until one or more model quality criteria are met, until a predetermined maximum number of iterations are performed, or until human intervention terminates the training process. Once the iterations of operations 420-440 are complete, the settlement predictor 102 may assemble the machine learning models 455 in operation 430, and the machine learning models 455 may be used for testing.

[0040] Once the machine learning model 455 is prepared for testing, the settlement predictor 102 may apply the machine learning model 455 to the test data set 460 in operation 470 to predict the probability of failure and the most likely cause of failure for each financial transaction represented in the test data set 460. In operation 470, the settlement predictor 102 may average or otherwise aggregate the predictions of the machine learning model 455 to generate a final prediction 490 for the test data set 460.

[0041] The final predictions 490 of the test data set 460 may be used to select a machine learning model 455 for producing financial transaction settlement predictions, or for further selecting and tuning model parameters. Figure 5 The performance of a machine learning model in a system for transaction settlement prediction according to one or more embodiments is depicted as a receiver operating characteristic (ROC) curve graph 500. Figure 5 As shown, ROC curve graph 500 can plot the model performance of the discrimination threshold settings that produce a combination of true positive rate 510 and false positive rate 520. Curve 540 represents the performance of a single model 455 or an aggregate model 455. Point 560 represents a true positive rate of 0.5, and point 560 represents the best performing discrimination threshold. Curve 530 representing pure chance prediction is provided as a reference.

[0042] Figure 6 1 is a flow chart of a method for transaction settlement prediction according to one or more embodiments. In operation 610, settlement predictor 102 may receive data of past financial transactions, such as data from database 130. In operation 615, settlement predictor 102 may Figure 4The illustrated process and the processes discussed in detail above train machine learning models using data from past financial transactions. In operation 620, settlement predictor 102 may receive parameters of a recently executed subject financial transaction. In operation 630, settlement predictor 102 may determine a likelihood that settlement of the transaction will fail using the trained machine learning model. In operation 640, settlement predictor 102 may determine the most likely reason why settlement of the transaction will fail using the trained machine learning model. In operation 650, settlement predictor 102 may display a message in a user interface (e.g., as described below in conjunction with Figure 7-9 The likelihood and most likely reason that the transaction settlement will fail may be displayed on a user interface (discussed in the example of FIG. 6 ). In operation 660, the settlement predictor 102 may pause for a predetermined length of time. In operation 670, the settlement predictor 102 may determine whether a recently executed financial transaction has been settled. The pause before re-evaluating the likelihood of a settlement failure may be determined automatically based on the currently assessed likelihood, general user settings, or a specific time interval set by the user for this transaction. If the recently executed financial transaction has not been settled, the settlement predictor 102 may return to operation 630. If the recently executed financial transaction has been settled, in operation 680, the settlement predictor 102 may add this transaction to the data of past financial transactions (e.g., database 130) and then return to operation 620 to retrain the model based on the updated database 130. Retraining of the model may be performed according to a set schedule (i.e., weekly or monthly), may be performed based on performance degradation of the trained model, and may be manually initiated.

[0043] As described above, the settlement predictor 102 can display the probability and most likely reason why a transaction will fail on the user interface, such as Figure 7-9 shown.

[0044] Figure 7 A user interface 108 for presenting a probability of transaction failure according to one or more embodiments is depicted. Figure 7 As shown, user interface 108 may present a representation 740 of financial transactions ranked according to probability of failure 720 and number of days until expected settlement 730. For example, financial transactions may be classified as having high, medium, or low probability of failure, as shown on axis 720. In addition, financial transactions for a current settlement date, a next settlement date, and a second settlement date may also be depicted. Each financial transaction may be depicted using a graphical element, such as a circle (e.g., Figure 7) or other shapes. The size of each financial transaction representation 740 may correspond to the relative risk of failure of the financial transaction, the number or amount of the financial transaction, or a risk-weighted measure to indicate the relative importance of the financial transaction. The metric corresponding to the size of each financial transaction representation 740 may be selected based on user preferences. In addition, the financial transaction representations 740 may be differentiated based on other visual attributes (such as color or shading) to indicate whether the financial transaction is a "sell" transaction or a "buy" transaction, etc. (such as Figure 7 If the user selects a financial transaction representation 740, the user interface may display financial transaction details 750, including the calculated probability that the financial transaction will fail and the most likely reason for the failure, as well as other attributes of the financial transaction (transaction, date, security traded, buy or sell, etc.).

[0045] Figure 8 A user interface 805 is depicted that displays information about risk transaction settlement (such as information that can be determined by the above method) and automatically generated opinions on risk transaction settlement according to one or more embodiments. Figure 8 As shown, the user interface 805 may include a graphical pane 810 and a text pane 815. The graphical pane 810 may include one or more graphical elements that display summary data about current risk trade settlements, such as data that can be determined by the above method. For example, the graphical pane 810 may include a linear graph 820 of the number of current risk trade settlements over time, a linear graph 825 of the cumulative value of current risk trade settlements over time, a bar graph 830 of current risk trade settlements divided by counterparties, and a bar graph 835 of current risk trade settlements divided by reasons. The selection of graphical elements displayed in the graphical pane 810 and the order in which they appear may be determined by user preferences. The text pane 815 may display a text description of current trends and other views on risk trade settlements. The current trends and other views displayed in the text pane 815 may be automatically generated, which may be specific to a user or general to the entire organization.

[0046] Fig. 9 A user interface 905 is depicted according to one or more embodiments, the interface displaying information about risk transaction settlement (such as information that can be determined by the above method) and detailed information about a selected risk transaction settlement. Fig. 9 As shown, user interface 905 may include a table pane 965 of risk transaction settlements and a details pane 975 for detailed information about a risk transaction settlement selected from table pane 965. Table pane 965 may include one or more filter elements 910 for selecting a subset of risk transaction settlements to be displayed. For example, the risk transaction settlements may be displayed by mitigation status (e.g., Fig. 9) or any other attribute of the risk transaction settlement. The available filters may be determined by user preferences, which may be specific to a user or may be generally determined for an organization. The table pane 965 may include a row 970 for each risk transaction settlement that meets the criteria set by the filter element 910. Each row 970 may include detailed information about each risk transaction settlement, such as a weighted risk score 915, a raw risk score 920, a mitigation status 925, a transaction execution date 930, a settlement date 935, a security identifier 940, a transaction type 945 (such as a buy or sell), a transaction net amount 950, a reason why the transaction may fail to settle 955, and a counterparty 960. For example, the raw risk score 920 may include a direct / automatic output of a predictive model. In addition, the weighted risk score 915 may optionally include an institution-specific factor, which may be a weighting. For example, such a weighting may be subject to human influence by the user and / or institution. For example, institution-specific factors may include the importance of certain clients, knowledge of certain (recent) high-risk markets, specific weighting of transaction types, or over-weighting of certain scenarios, etc., which may be taken into account, for example, in cases where the raw risk score 920 provides ambiguous results. The detailed formation items displayed in each row 970 may be determined by user preferences, which may be specific to a particular user or may be generally determined for an organization. If a user selects a row 970, the user interface 905 may display a details pane 975 with information about the selected risky transaction. Details pane 975 may display detailed information, which may include the same information displayed in table pane 965, or may display more or less information based on user preferences. If a user selects an item of information in details pane 975, the user may edit the displayed information or other settings related to the selected risky transaction. For example, if a user selects an item of information in details pane 975, the user may edit the displayed information or other settings related to the selected risky transaction. Fig. 9 As shown, if the user selects the "Mitigate" item, the user can change the current mitigation status of the selected risk transaction.

[0047] To implement transaction settlement predictions, any suitable system infrastructure may be established. Figure 1A and 1B The foregoing discussion briefly and generally describes a suitable computing environment in which the present disclosure may be implemented. In one embodiment, any disclosed system, method, and / or graphical user interface may be implemented by a computer system that is compatible with the present disclosure. Figure 1A and 1BThe system shown in the figure is executed or implemented by a computing system consistent with or similar to the system shown. Although not required, various aspects of the present disclosure are described in the context of computer-executable instructions (e.g., routines executed by data processing devices (such as server computers, wireless devices, and / or personal computers)). It will be understood by those skilled in the relevant art that various aspects of the present disclosure can be practiced through other communication, data processing, or computer system configurations, including: Internet appliances, handheld devices (including personal digital assistants ("PDAs")), wearable computers, various cellular phones or mobile phones (including voice transmission over IP ("VoIP")), dumb terminals, media players, gaming devices, virtual reality devices, multi-processor systems, microprocessor-based or programmable consumer electronics, set-top boxes, network PCs, microcomputers, mainframe computers, etc. In fact, in the present invention, the terms "computer", "server", etc. are generally used interchangeably to refer to any of the above devices and systems and any data processors.

[0048] Various aspects of the present disclosure may be embodied in special-purpose computers and / or data processors that are specially programmed, configured, and / or constructed to execute one or more computer-executable instructions detailed in the present invention. Although various aspects of the present disclosure (such as certain functions) are described as being executed entirely on a single device, the present disclosure may also be practiced in a distributed environment in which different processing devices share functions or modules that are connected via a communication network (such as a local area network ("LAN"), a wide area network ("WAN"), and / or the Internet). Similarly, the techniques described in the present invention involving multiple devices may be implemented in a single device. In a distributed computing environment, program modules may be located simultaneously in local and / or remote memory storage devices.

[0049] Various aspects of the present disclosure may be stored and / or distributed on non-transitory computer readable media (including magnetic or optical readable computer disks, hardwired or pre-programmed chips (such as EEPROM semiconductor chips), nanotechnology memories, biological memories, or other data storage media). Alternatively, over a period of time, computer-implemented instructions, data structures, screen displays, and other data under various aspects of the present disclosure may be distributed on the Internet and / or other networks (including wireless networks), on propagation signals of propagation media (such as electromagnetic waves, sound waves, etc.), and / or may be provided on any analog or digital network (packet switching network, circuit switching network or other scheme).

[0050] Other embodiments of the present disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. The specification and examples are intended to be exemplary only, with the true scope and spirit of the present invention being indicated by the following claims.

Claims

1. A computer-implemented method for predicting transaction settlement, the method comprising: receiving data of a plurality of past financial transactions; training a machine learning model using data from the plurality of past financial transactions; receiving one or more parameters of a subject financial transaction from among a plurality of recently executed financial transactions; Using trained machine learning models to determine the likelihood that a subject financial transaction will fail; Using trained machine learning models to determine the most likely reasons why a subject financial transaction will fail; The user is presented with a likelihood that the subject financial transaction will fail and the most likely reason why the subject financial transaction will fail.

2. The computer-implemented method of claim 1 , wherein: The likelihood and most likely cause are presented via a user interface that includes a user interface element for each of a plurality of recently executed financial transactions, a size of each user interface element representing a likelihood that the respective financial transaction will fail, a number of the respective financial transactions, an amount of the respective financial transaction, or a risk-weighted measure of the importance of the respective financial transaction relative to other financial transactions in the plurality of financial transactions.

3. The computer-implemented method of claim 1 , further comprising: updating the data of the plurality of past financial transactions by adding the subject financial transaction, the likelihood, the most likely cause, the failure or settlement date of the subject financial transaction, and the failure cause of the subject financial transaction to the data of the plurality of past financial transactions; Retrain machine learning models using updated data from past financial transactions.

4. The computer-implemented method of claim 3, wherein: After a pause of a predetermined length of time, the machine learning model is retrained using updated data from past financial transactions.

5. The computer-implemented method of claim 1 , further comprising: Pause for a specified period of time; Using a trained machine learning model to update the probability that a subject financial transaction will fail; The most likely reasons why updating the subject financial transaction using a trained machine learning model will fail.

6. The computer-implemented method of claim 5, wherein: The specified time period is automatically determined based on the probability, or is a predetermined value.

7. The computer-implemented method of claim 1 , further comprising: Tune the performance of a machine learning model based on user-specified tuning parameters.

8. A system for transaction settlement prediction, the system comprising: a data storage device for storing instructions for transaction settlement forecasts in an electronic storage medium; The processor can execute the instructions to perform a method comprising the following steps: receiving data of a plurality of past financial transactions; training a machine learning model using data from the plurality of past financial transactions; receiving one or more parameters of a subject financial transaction from among a plurality of recently executed financial transactions; Using trained machine learning models to determine the likelihood that a subject financial transaction will fail; Using trained machine learning models to determine the most likely reasons why a subject financial transaction will fail; The user is presented with a likelihood that the subject financial transaction will fail and the most likely reason why the subject financial transaction will fail.

9. The system according to claim 8, wherein: The likelihood and most likely cause are presented via a user interface that includes a user interface element for each of a plurality of recently executed financial transactions, a size of each user interface element representing a likelihood that the respective financial transaction will fail, a number of the respective financial transactions, an amount of the respective financial transaction, or a risk-weighted measure of the importance of the respective financial transaction relative to other financial transactions in the plurality of financial transactions.

10. The system according to claim 8, wherein: The system may further: updating the data of the plurality of past financial transactions by adding the subject financial transaction, the likelihood, the most likely cause, the failure or settlement date of the subject financial transaction, and the failure cause of the subject financial transaction to the data of the plurality of past financial transactions; Retrain machine learning models using updated data from past financial transactions.

11. The system according to claim 10, wherein: After a pause of a predetermined length of time, the machine learning model is retrained using updated data from past financial transactions.

12. The system according to claim 8, wherein: The system may further: Pause for a specified period of time; Using a trained machine learning model to update the probability that a subject financial transaction will fail; The most likely reasons why updating the subject financial transaction using a trained machine learning model will fail.

13. The system according to claim 12, wherein: The specified time period is automatically determined based on the probability, or is a predetermined value.

14. The system according to claim 8, wherein: The system may further: Tune the performance of a machine learning model based on user-specified tuning parameters.

15. A non-transitory machine-readable medium storing instructions, which, when executed by a computing system, causes the computing system to perform a transaction settlement prediction method, the method comprising: receiving data of a plurality of past financial transactions; training a machine learning model using data from the plurality of past financial transactions; receiving one or more parameters of a subject financial transaction from among a plurality of recently executed financial transactions; Using trained machine learning models to determine the likelihood that a subject financial transaction will fail; Using trained machine learning models to determine the most likely reasons why a subject financial transaction will fail; The user is presented with a likelihood that the subject financial transaction will fail and the most likely reason why the subject financial transaction will fail.

16. The non-transitory machine-readable medium of claim 15, wherein: The likelihood and most likely cause are presented via a user interface that includes a user interface element for each of a plurality of recently executed financial transactions, a size of each user interface element representing a likelihood that the respective financial transaction will fail, a number of the respective financial transactions, an amount of the respective financial transaction, or a risk-weighted measure of the importance of the respective financial transaction relative to other financial transactions in the plurality of financial transactions.

17. The non-transitory computer readable medium of claim 15, the method further comprising: updating the data of the plurality of past financial transactions by adding the subject financial transaction, the likelihood, the most likely cause, the failure or settlement date of the subject financial transaction, and the failure cause of the subject financial transaction to the data of the plurality of past financial transactions; Retrain machine learning models using updated data from past financial transactions.

18. The non-transitory machine-readable medium of claim 17, wherein: After a pause of a predetermined length of time, the machine learning model is retrained using updated data from past financial transactions.

19. The non-transitory computer readable medium of claim 15, the method further comprising: Pause for a specified period of time; Using a trained machine learning model to update the probability that a subject financial transaction will fail; The most likely reasons why updating the subject financial transaction using a trained machine learning model will fail.

20. The non-transitory machine-readable medium of claim 19, wherein: The specified time period is automatically determined based on the probability, or is a predetermined value.

Images