Process flow event sequence prediction

Abstract

It provides process flow event sequence prediction. [Solution] According to one aspect of one embodiment, the operation may include generating a first representation of a first event in a process flow corresponding to an incident. A second representation of a second event, which may be a prediction of a future event following the first event in the process flow, may be generated using an artificial intelligence model based on the first representation. A third representation of a third event, which may be a prediction of a future event following the second event in the process flow, may be generated using an artificial intelligence model based on the first and second representations. A predicted event sequence for the entire process flow may be generated based on the predicted order of the first, second, and third events.

Description

TECHNICAL FIELD

[0001] Embodiments discussed herein relate to process flow behavior prediction, including process flow prediction based on partial flow information using machine learning, for example.

BACKGROUND ART

[0002] Process flows are used across various industries and domains to describe the sequence of events associated with a particular incident. In some examples, process flows are designed to represent tasks and operations aimed at achieving a goal. These flows provide a structured framework for capturing and organizing events and interactions within a process, enabling clear communication and analysis. While process flows can also guide task execution, their descriptive nature allows organizations to document and understand the progression of events, decision points, and outcomes related to a specific scenario or incident. Process flows are often visualized using diagrams, charts, or text descriptions that show each step, decision point, and their interconnections, facilitating streamlined operations, error reduction, and productivity improvement.

[0003] The subject matter claimed herein is not limited to embodiments that solve any disadvantages or that operate only in environments such as those described above. Rather, this background is only provided to illustrate one exemplary technical field in which some embodiments described herein may be implemented.

SUMMARY OF THE INVENTION

MEANS FOR SOLVING THE PROBLEMS

[0004] According to one aspect of one embodiment, the operation may include generating a first representation of a first event in a process flow responding to an incident. A second representation of a second event, which may be a prediction of a future event following the first event in the process flow, may be generated using an artificial intelligence model based on the first representation. A third representation of a third event, which may be a prediction of a future event following the second event in the process flow, may be generated using an artificial intelligence model based on the first and second representations. The predicted event sequence for the entire process flow may be generated based on the predicted order of the first, second, and third events.

[0005] The objectives and advantages of the embodiments are realized and achieved, at least, by the elements, features, and combinations specifically indicated in the claims. It is understood that the above general description and the following detailed description are illustrative and descriptive, and not limiting to the claimed invention. [Brief explanation of the drawing]

[0006] Exemplary embodiments are described and explained with further specificity and detail through the use of the accompanying drawings.

[0007] [Figure 1] This disclosure describes an environment related to generating a predicted complete process flow based on partial process flow data, according to one or more embodiments of this disclosure.

[0008] [Figure 2] This disclosure illustrates an exemplary system relating to generating one or more event embeddings based on partial process flow data, according to one or more embodiments of this disclosure.

[0009] [Figure 3]This disclosure provides an exemplary system relating to generating one or more predicted event representations, according to one or more embodiments of this disclosure.

[0010] [Figure 4] This disclosure provides an exemplary system relating to extracting one or more predicted features from a predicted feature representation, according to one or more embodiments of this disclosure.

[0011] [Figure 5] An exemplary flowchart of an exemplary method for generating a predicted event sequence for a process flow, according to one or more embodiments of the present disclosure, is shown.

[0012] [Figure 6] A block diagram of an exemplary computing system according to at least one embodiment of the present disclosure is shown. [Modes for carrying out the invention]

[0013] Many organizations and systems use process flows, for example, to describe the progression of events or activities related to a particular incident. By capturing the temporal or event-driven dynamics of a process, process flows provide organizations with a descriptive framework for modeling and visualizing workflows, identifying interdependencies, gaining insights into behavioral patterns, and taking action based on process flows. Widely adopted across industries, process flows enable organizations to track how an incident unfolds over time, providing clarity to complex systems without prescribing rigid execution paths. Due to this versatility, process flows become crucial for improving transparency, resolving incidents, and adapting to the unique needs of diverse operating environments.

[0014] This disclosure relates to a system and method for predicting process flow sequences and corresponding timings using artificial intelligence. In particular, some embodiments may include techniques for generating predicted event sequences about process flows using deep learning models.

[0015] Aspects of this disclosure provide a system for predicting future events in a process flow. The system may include an event representation module configured to generate representations of events (e.g., event embeddings) in a partial event log of an ongoing process. In these and other embodiments, a generative artificial intelligence (AI) model (e.g., configured as an autoregressive transformer decoder) may be used to process the event representations and generate contextual representations of predicted future events in the process flow. The system may further use a classification head to predict specific attributes of future events based on the contextual representations.

[0016] By leveraging deep learning techniques such as transformer models, the disclosed system may be capable of capturing long-range dependencies in event sequences and generating accurate predictions of how the process might unfold. This could help enable proactive optimization of the process by predicting potential problems or bottlenecks before they occur. The system can also predict attributes such as the next activity, event duration, required resources, and overall process completion time.

[0017] The decoder's autoregressive nature allows the system to not only predict the next single event, but also to sequentially and iteratively generate a complete predicted event sequence for the rest of the process flow. This predicted event sequence provides a holistic view of the expected process progression, which can be analyzed to detect potential deviations, deadlocks, or process restarts.

[0018] In these and other embodiments, based on this analysis, the system may generate recommendations for resource allocation or process modification to improve efficiency. Resource allocation and / or process modification based on the predicted event sequence can be implemented in a variety of ways.

[0019] For example, the system may automatically adjust staffing levels or schedules based on the predicted workload. If the predicted event sequence indicates a large number of incidents expected during a certain time period, additional personnel may be scheduled for that time frame.

[0020] In these and other embodiments, resource allocation may be optimized based on predicted event types and durations. For example, if the system anticipates that certain types of complex problems are likely to occur, experts with relevant expertise may be pre-assigned.

[0021] Additionally or alternatively, process steps may be reordered or parallelized to improve efficiency. In these and other embodiments, if the predicted sequence suggests a potential bottleneck, the workflow may be modified to distribute the work more evenly.

[0022] By proactively adjusting resources and processes based on the predicted event sequences, the system can improve the efficiency of process flow completion and the use of resources for executing process flows. Specific adjustments can be tailored to the unique needs and constraints of each business process.

[0023] The training process for a generative AI model (the "model") may include several steps to improve its ability to perform process flow prediction. First, the model may be trained on a large dataset of historical event logs from completed process instances. This dataset may include sequences of events with relevant attributes such as activity names, timestamps, resources, and other relevant features.

[0024] During training, the model can learn to capture temporal dependencies and patterns within those event sequences through its self-attention mechanism. Multi-head attention layers can allow the model to focus on different aspects of the input sequence when making predictions.

[0025] The training process may also utilize techniques such as masked language modeling. Here, some events in the input sequence are randomly masked, and the model is tasked with predicting those masked events. This can help improve the model's ability to learn bidirectional context and fill in missing information.

[0026] The model can also be trained simultaneously for multiple prediction tasks, such as predicting the next activity, timestamp, resource allocation, and other relevant attributes. This multi-task learning approach can enable the model to develop a more comprehensive understanding of process dynamics.

[0027] To improve generalization, training data may include a diverse set of process variations and edge cases. Techniques such as data augmentation may be used to artificially increase the diversity of training examples.

[0028] The model's performance can be evaluated during training using a provided validation set. Hyperparameters such as the learning rate, batch size, and model architecture can be tuned based on the validation performance.

[0029] Through sequential, iterative training and refinement, generative AI models can develop the ability to effectively capture long-term dependencies in event sequences, represent complex event attributes, and generate accurate predictions for future process behavior. This improved predictive capability can enable more effective online process monitoring and optimization.

[0030] Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0031] Figure 1 shows an environment 100 related to the generation of a predicted complete process flow 120 based on partial process flow data 102 according to one or more embodiments of the present invention.

[0032] Partial process flow data 102 may include any data associated with a process flow that has not yet been completed. As used in this disclosure, a process flow is a description of events corresponding to an incident. References to “incident” in this disclosure may refer to any category of events, conditions, or situations that can be resolved using an action or intervention. In these and other embodiments, an incident may include a particular situation or project that may include several different events. For example, in the context of customer service, an incident may include a customer device that may not be functioning properly. Continuing this example, the customer device may have several anomalies that require service (e.g., a broken screen, a missing or broken sensor). In such an example, the overall incident—a malfunctioning device—may include a series of events that occur as part of repairing or replacing the device to resolve such an incident. Additionally or alternatively, an incident may refer to a single situation or event. For example, in the context of pedestrian transport using scooters, an incident may include a single ride on a scooter from a starting point to a destination point. As an additional example, an incident may include a car rental, where the incident may include any events, changes in state, or activities that may occur from the time the rental car is picked up until it is dropped off.

[0033] In some embodiments, the partial process flow data 102 may include data associated with unresolved incidents that have not yet been completed.

[0034] In some embodiments, the partial process flow data 102 may further include data associated with an event or activity corresponding to an incident. In these and other embodiments, such an event may have already been completed. The event or activity may include any action, change, or occurrence that takes place in relation to the incident. In some embodiments, the event may include an action taken to address the incident, such as a diagnostic test, corrective repair, or stakeholder communication.

[0035] In some embodiments, events and / or activities associated with an incident may move the incident toward or toward resolution. Additionally or alternatively, activities and / or events may not affect whether an incident moves toward or toward resolution; instead, they may describe events that may have occurred in connection with the incident.

[0036] For example, an incident may involve a technical problem related to a software program. Continuing this example, events or activities associated with an incident may include performing system diagnostics, applying software patches, and / or replacing faulty hardware components.

[0037] As an additional example, an incident could include customer-related issues, such as a problem a customer may have raised during a call with an agent. Continuing this example, an event and / or activity could include contacting the customer for additional details, escalating the issue to a specialist team, and / or issuing a refund.

[0038] In some embodiments, events may further include state changes. State changes may indicate transitions in the status or state of an incident. In some embodiments, state changes can provide a high-level view of progress, such as moving from an “open” state to an “in progress” or “resolved” state. For example, a technical issue initially flagged as “unresolved” may transition to “diagnosed” after root cause analysis, and then to “resolved” once a fix is ​​implemented. Similarly, in another example, a customer complaint may progress from “submitted” to “under consideration” and then “closed” after the issue has been addressed to the customer's satisfaction.

[0039] In some embodiments, a state change may be accompanied by one or more events. Additionally or alternatively, a state change may not be accompanied by any events. For example, an incident may transition from a “pending” state to a “closed” state due to a lack of activity.

[0040] While some embodiments have been described with reference to exemplary industries, completed process flow data 102, including, for example, incidents, events, activities, and state changes, can occur in a variety of industries, domains, and / or environments, including, but not limited to, technical systems, business processes, security contexts, and customer interactions, to name a few.

[0041] In some embodiments, the partial process flow data 102 may include a partial event log. The partial event log may include various attributes (also referred to in this disclosure as “features”) for each event shown in the partial process flow data 102. In some embodiments, these attributes may include an event identifier, an activity name, a resource identifier, a timestamp, and / or other relevant information.

[0042] An event identifier may be a unique alphanumeric code or label assigned to each specific event within a process flow. This identifier serves to distinguish individual events and allows for easy referencing and tracking within the system.

[0043] Resources can be any entity, such as people, teams, software applications, or hardware components, that are used to perform activities within a process flow. Resources can be assigned to specific events or tasks within a process.

[0044] An activity may involve a specific task, action, or step performed as part of a corresponding event. An activity may represent a discrete unit of work that contributes to moving a process toward completion. In these and other embodiments, an activity may indicate a different state change or status in a process flow corresponding to a particular event.

[0045] Comments may include additional text information or notes associated with a specific event or activity within a process flow. Comments can provide context, explanation, or other relevant details about the event.

[0046] A timestamp can correspond to the date and time a particular event or activity occurred or was recorded within a process flow. Timestamps can allow for the temporal ordering and temporal analysis of events within a process.

[0047] Event logs can potentially capture the progress of a process over time, spanning multiple days. For example, a log might record events that occurred from day 1 to day 9 of process execution. The sequence of events might include stages showing certain activities associated with discrete events, such as:

[0048] Day 1: Open

[0049] Day 2: Registration complete

[0050] Days 3-4: Survey

[0051] Days 5-6: Revisions are underway.

[0052] Day 7: On hold - Assigned to a third party

[0053] Day 8: Action Needed

[0054] Day 9: End

[0055] Event logs can also capture branching paths and loops within a process flow. For example, if further investigation is needed after the "Action Required" stage, the process may loop back to "Investigation." Alternatively, depending on the specific circumstances of the case, it may proceed to "Correction in Progress" or directly to "Close."

[0056] In some embodiments, partial process flow data 102 may be processed by an event representation module 104. The event representation module 104 may be implemented using hardware including one or more processors, a central processing unit (CPU), a graphics processing unit (GPU), a data processing unit (DPU), a parallel processing unit (PPU), a microprocessor (e.g., one or more operations that perform or control the execution thereof), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), an accelerator (e.g., a deep learning accelerator (DLA)), and / or other processor types. In these and other embodiments, the event representation module 104 may be implemented using a combination of hardware and software.

[0057] Additionally or alternatively, the event representation module 104 may include any suitable artificial intelligence (AI) model and / or may be implemented using any suitable AI model. Some examples of AI models that may be incorporated include: recurrent neural networks (RNNs) such as long short-term memory (LSTM) networks or gated recurrent units (GRUs); transformer models such as BERT (Bidirectional Encoder Representations from Transformers) or GPT® (Generative Pre-trained Transformer); convolutional neural networks (CNNs); random forests; support vector machines (SVMs); gradient boosting models such as XGBoost or LightGBM; hidden Markov models; conditional random fields; Bayesian networks; Gaussian processes; deep belief networks; autoencoders; generative adversarial networks (GANs); reinforcement learning models; and ensemble methods that combine multiple model types. A specific AI model architecture can be selected and customized based on the specific task, available data, and desired performance characteristics. In some implementations, multiple model types may be combined. AI models can be trained on historical process data to learn patterns and make predictions about future process behavior.

[0058] In this disclosure, the actions described as being performed by the event representation module 104 may include actions that the event representation module 104 can instruct a corresponding computing system to perform. In these or other embodiments, the event representation module 104 may be implemented and / or run by one or more computing systems, such as those described in more detail with respect to Figure 6.

[0059] The event representation module 104 may be configured to generate one or more event representations 106 based on partial process flow data 102. In these and other embodiments, the event representation module 104 may generate an event representation 106 for each event shown in the partial process flow data 102.

[0060] The event representation 106 can generally represent different characteristics of an event. In some embodiments, such a representation may be generated as an event embedding that includes a combination of individual feature embeddings. Each individual feature embedding may be generated for different attributes or characteristics of the event. For example, the event representation 106 may include embeddings for attributes such as activity type, resource, timestamp, case identifier, or other event metadata.

[0061] Individual feature embeddings can be combined in various ways to generate an overall event embedding. In some implementations, feature embeddings can be concatenated to form an event embedding. For example, for an event with feature embeddings e1, e2, and e3 for three different corresponding attributes, the overall event embedding may be [e1;e2;e3], where ";" indicates concatenation.

[0062] In these and other embodiments, feature embeddings may be combined through element-wise operations such as addition or multiplication. Event embeddings may be generated by passing individual feature embeddings through a neural network layer to learn the optimal combination. Various pooling operations, such as max pooling or mean pooling, may also be applied to combine feature embeddings.

[0063] In these and other embodiments, the event representation 106 may be configured to show a sequence of events indicated by the partial process flow data 102. For example, timestamps associated with different events may be used to show a sequence of events, and the event representation 106 may be organized according to such a sequence.

[0064] The event representations 106 generated through these embedding techniques can capture the target characteristics of each event in a fixed-dimensional vector space. This allows subsequent components of the environment 100 to process and analyze the partial process flow data 102 more efficiently. Figure 2 shows further examples of event embedding generation that can be used as event representations 106 in some embodiments.

[0065] In some embodiments, the environment 100 may include a generative AI model 108 configured to analyze event representations 106 to generate one or more predicted event representations 110. The AI ​​model 108 may include any suitable AI model that can be configured to perform the operations of the generative AI model 108 described herein. For example, in some embodiments, the generative AI model 108 may be configured as an autoregressive transformer decoder, as described in more detail with respect to Figure 3.

[0066] The generative AI model 108 may receive an event representation 106 of an event contained in partial process flow data 102 and generate predictions of one or more future events and corresponding sequences of such future events. Such future events may be represented as predicted event representations 110. An example of a predicted event representation 110 may be a contextual representation of the next event that can provide information about the event, which may be useful in the implementation of a process flow and / or in the prediction of events and event sequences. For example, a predicted event representation 110 generated by the generative AI model 108 may capture a variety of attributes of the predicted future event. For example, a predicted event representation 110 may include predictions of the next activity, resource allocation, timestamp, duration, and other event characteristics.

[0067] For example, a predicted event representation 110 of a particular predicted event may include an event embedding. In an event embedding, the particular predicted event may be represented as a dense vector embedding that encodes information about the event's attributes and characteristics (also called "features"). This embedding may be generated by a generative AI model 108 based on the context of a previous event, such as that shown by the event representation 106.

[0068] Additionally or alternatively, in some embodiments, a particular predicted event representation 110 may include temporal information such as a predicted timestamp or duration of a particular event. This allows for the modeling of time-based aspects of the process flow.

[0069] In these and other embodiments, a particular predicted event representation 110 may include a sequential context for that particular predicted event. For example, a particular predicted event representation 110 incorporates contextual information from a sequence of previous events (indicated by event representation 106) to capture dependencies and patterns in the process flow. In these and other embodiments, the sequential context may be encoded using an attention mechanism or other sequential modeling techniques. Additionally or alternatively, the sequential context may include a positional encoding that indicates the position of a particular event in the overall sequence of events in the process flow. For example, in some embodiments, the positional encoding may provide information about where a particular event occurs relative to other events. Furthermore, in some embodiments, temporal information may be used and / or included in the sequential context information.

[0070] Additionally or alternatively, in some embodiments, a particular predictive event representation 110 may include task-specific information encoded therein. Task-specific information may include task-specific attributes relevant to that particular event and / or process flow, such as allocated resources, priority levels, or other domain-specific features that may correspond to that particular event and / or process flow.

[0071] In these and other embodiments, a particular predictive event representation 110 may include one or more probability distributions contained therein. For example, rather than a single-point prediction, a particular predictive event representation 110 may include one or more probability distributions across possible event features that capture uncertainty in the prediction.

[0072] Additionally or alternatively, in some embodiments, a particular predictive event representation 110 may include feature importance information. Feature importance information may indicate which input features (e.g., those represented by event representation 106) were most important or influential in generating the prediction for this particular event.

[0073] In these and other embodiments, a particular predicted event representation 110 may include one or more anomaly indicators. Anomaly indicators may include signals indicating that a particular predicted event appears anomaly or deviates significantly from an expected pattern based on historical data.

[0074] In some embodiments, the predicted event representation 110 may be combined with the event representation 106 as a recursive input to the generating AI model 108. Thus, the predicted events and the corresponding predicted event representations 110 may be used to predict other future events and generate other predicted event representations 110 and corresponding event sequences (e.g., predicted or already occurring) until a predicted complete process flow 120 is reached.

[0075] The predicted complete process flow 120 may include a prediction of the entire process flow from incident occurrence to resolution, corresponding to the partial process flow data 102. The predicted complete process flow 120 can predict various aspects of future events in the process. For example, the predicted complete process flow 120 may predict that the next event after Team C's "Investigation" on day 4 is "Correction in Progress" by Team C on day 8. It may also predict that the process will conclude with Team C's "Closing" on day 11. The predicted complete process flow 120 can also identify potential problems, such as predicting a loop between events 4, 5, and 6 (which may indicate a deadlock in the process). Furthermore, it may predict the possibility of the process resuming, such as returning to the initial "Open" event after event 4.

[0076] For example, a predicted complete process flow 120 can predict the sequence of future activities. Given a partial event log indicating that an incident has been opened and accepted, the generating AI model 108 can predict the next activities as "investigate," followed by "correction in progress," and finally "closed." The generating AI model 108 can also predict when each predicted activity may occur.

[0077] Therefore, in some embodiments, the generating AI model 108 can perform multitask predictions to generate various characteristics of predicted future events and process flows. As described above, these multitask predictions may include predicting the next activity, event duration, remaining time to process completion, potential deviations, deadlocks, or process restarts. In some embodiments, the predicted complete process flow 120 may include predicted event representations 110 that can be ordered or specified according to the order of corresponding events in the predicted complete process flow 120.

[0078] In some embodiments, the generative AI model 108 includes and / or can be implemented as an autoregressive decoder-only transformer model. The autoregressive decoder-only transformer model may employ a multi-head self-attention mechanism to capture long-term dependencies in the event sequence in order to obtain the predicted complete process flow 120. Further details regarding the generative AI model 108 implemented as an autoregressive transformer decoder model are given with respect to Figure 3.

[0079] Therefore, the generative AI model 108 can incorporate composite information about each event (for example, as shown by the event representation 106) in generating predictions. As described above, this composite information may include multiple attributes of each event, such as the activity name, timestamp, assigned resources, work notes, comments, or other event metadata.

[0080] The predicted complete process flow 120, generated by sequentially and iteratively applying the generative AI model 108, can enable analysis of the complete predicted sequence. This analysis can allow for the detection of potential problems in the predicted future process flow, such as undesirable deviations, bottlenecks, or inefficiencies.

[0081] The generative AI model 108 can be trained to generate a predicted complete process flow 120 based on complete process flow data corresponding to a completed process flow, using partial process flow data 102. The training process may involve feeding the complete process flow sequence to the generative AI model 108 and training the model to predict subsequent events given partial sequences as input.

[0082] For example, in some embodiments, during training, the generative AI model 108 may be given prefixes of events from a complete process flow and tasked with predicting the rest of the sequence. The generative AI model 108's predictions may then be compared to the actual subsequent events to calculate the loss and update the generative AI model 108's parameters. This allows the generative AI model 108 to learn patterns and relationships between events in the process flow.

[0083] In some embodiments, the training data may include a large corpus of complete process flows from various processes. One or more training examples may include a partial sequence of events as input and the corresponding complete sequence as the target output. The generative AI model 108 may learn to map from partial sequences to complete sequences through many iterations on the training data.

[0084] The training process for the generative AI model 108 may include various aspects in several embodiments. For example, diverse attributes of each event in the completed process flow data may be embedded in a unified vector representation. Masked self-attention may be used in the transformer architecture to condition predictions for partial sequences provided as input. In these and other embodiments, teacher forcing may be utilized so that the generative AI model 108 is given correct previous events when predicting each subsequent event during training. Additionally or alternatively, techniques such as curriculum learning may be applied, where the generative AI model 108 is initially trained on shorter sequences and gradually exposed to longer and more complex process flows. In these and other embodiments, regularization techniques such as dropout may be used to improve generalization to unknown process flows.

[0085] By learning from a large dataset of complete process flows, the generative AI model 108 develops the ability to infer likely future events and patterns given an incomplete sequence. Thus, the trained generative AI model 108 can take partial process flow data 102 as input, generate predictions about subsequent events to complete the event sequence, and generate the predicted process flow 120.

[0086] In some embodiments, the environment 100 may include a feature extraction module 114. The feature extraction module 114 may be implemented using hardware including one or more processors, a central processing unit (CPU), a graphics processing unit (GPU), a data processing unit (DPU), a parallel processing unit (PPU), a microprocessor (e.g., one or more operations that perform or control the execution thereof), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), an accelerator (e.g., a deep learning accelerator (DLA)), and / or other processor types. In these and other embodiments, the feature extraction module 114 may be implemented using a combination of hardware and software.

[0087] Additionally or alternatively, the feature extraction module 114 may include and / or be implemented using any suitable artificial intelligence (AI) model. Some examples of AI models that may be incorporated include: recurrent neural networks (RNNs) such as long short-term memory (LSTM) networks or gated recurrent units (GRUs); transformer models such as BERT (Bidirectional Encoder Representations from Transformers) or GPT® (Generative Pre-trained Transformer); convolutional neural networks (CNNs); random forests; support vector machines (SVMs); gradient boosting models such as XGBoost or LightGBM; hidden Markov models; conditional random fields; Bayesian networks; Gaussian processes; deep belief networks; autoencoders; generative adversarial networks (GANs); reinforcement learning models; and ensemble methods that combine multiple model types. A specific AI model architecture can be selected and customized based on the specific task, available data, and desired performance characteristics. In some implementations, multiple model types may be combined. AI models can be trained on historical process data to learn patterns and make predictions about future process behavior.

[0088] In this disclosure, operations described as being performed by the feature extraction module 114 may include operations that the feature extraction module 114 can be instructed to perform on a corresponding computing system. In these or other embodiments, the feature extraction module 114 may be implemented and / or run by one or more computing systems, such as those described in more detail with respect to Figure 6.

[0089] The feature extraction module 114 may be configured to extract each predicted feature 116 corresponding to individual events represented by the predicted event representation 110 (for example, individually, output by the generative AI model 108 and / or included in the predicted complete process flow 120). In these and other embodiments, the predicted feature 116 may represent various features of the predicted event, such that the predicted feature is similar to or analogous to features included in the already completed event represented by the partial process flow data 102. Figure 4 illustrates and provides an exemplary process that may be performed to extract the predicted feature 116.

[0090] In some embodiments, the predicted complete process flow 120 and / or predicted features 116 (commonly referred to as “predictions”) may be communicated to or otherwise acquired by system 112. System 112 may include any suitable system, apparatus, or device configured to receive or otherwise acquire predictions and to perform one or more actions based on the predictions. In some embodiments, system 112 may include an event representation module 104 and / or a generative AI model 108. Additionally or alternatively, system 112 may be separated from one or more of the event representation module 104 or the generative AI model 108.

[0091] In some embodiments, system 112 may be configured to perform one or more actions based on predictions. For example, in the context of a customer service ticket as an ongoing process, the predictions may include data and / or information indicating the sequence of events that may need to occur to resolve the ticket. Based on these predictions, system 112 may automatically allocate predicted resources for individual events over a predicted period of time, in accordance with those predictions.

[0092] As another example, in the context of an ongoing process flow describing travel by rental car, the forecast may include data and / or information that may indicate events that may need to occur in order to successfully carry out the car rental. In response, system 112 may proactively adjust the rental schedule and allocate resources (e.g., personnel, vehicles, accessories, etc.) according to the predicted sequence of events and the corresponding characteristics indicated by the forecast.

[0093] In some embodiments, these predictions enable System 112 to make informed decisions, improve resource allocation, and improve customer satisfaction by predicting problems before they occur, better planning when and what events may occur, and identifying what may be required to execute and complete such events. Furthermore, System 112 can trigger automated workflows such as initiating follow-up actions, adjusting timelines, or notifying relevant parties, thus enabling the smooth and efficient execution of process flows.

[0094] Without departing from the scope of this disclosure, modifications, additions, or omissions may be made to the environment 100. For example, the order in which operations are performed by the various modules and components may be rearranged or performed in parallel in some embodiments. Additionally or alternatively, the event representation module 104 and the generative AI model 108 may be combined into a single module in certain implementations.

[0095] Furthermore, certain elements shown in Figure 1 may be divided into multiple subcomponents or combined in different ways. For example, the predicted event representation 110 and the predicted complete process flow 120 may both be generated by separate prediction modules rather than originating from the generating AI model 108. In some variations of the system, partial process flow data 102 may be input to the event representation module 104 after being preprocessed or filtered.

[0096] Figure 1 shows a specific arrangement of components and data flows; however, those skilled in the art will recognize that alternative configurations are possible to achieve similar functionality for predicting process flow completion and extracting relevant features. The core concepts of using event representation, generative models, and feature extraction can be implemented through various architectural designs without departing from the spirit of this disclosure.

[0097] Figure 2 shows an exemplary system 200 relating to generating one or more event embeddings 206 based on partial process flow data 202, according to one or more embodiments of the present disclosure. In these and other embodiments, the system 200 may relate to the operations performed by the event representation module 104 in Figure 1. Additionally or alternatively, the partial process flow data 202 may be similar to the partial process flow data 102 in Figure 1, and the event embeddings 206 may be the exemplary event representation 106 in Figure 1. For ease of explanation, the system 200 is described in relation to generating event embeddings 206 for a particular event, but the operations described may be performed to generate any number of corresponding event embeddings for any number of events.

[0098] The partial process flow data 202 may include event data 256 corresponding to a specific event. In these and other embodiments, the event data 256 may include feature data 250 that indicates features associated with a specific event.

[0099] In some embodiments, the system 200 may include an event representation module 204, which may be an example of the event representation module 104 in Figure 1. In these and other embodiments, the event representation module 204 may include an embedding layer 252 ("embedder 252"). The embedding layer may include a plurality of embedder components (embedder 1, embedder 2, ... embedder i, ... embedder n) for embedding different event features (event feature 1, event feature 2, ... event feature i, ... event feature n) represented by feature data 250 of the event data 256. In these and other embodiments, each embedder component may generate a feature embedding for its corresponding event feature.

[0100] The embedder 252 can operate in a manner involving multiple components and processes to generate feature embeddings for various individual features, such as event feature 1, event feature 2, etc. Each feature is fed to a corresponding embedder component, such as embedder 1, embedder 2, etc. These components are specifically designed to handle the particular type of feature data they receive.

[0101] For example, a lookup table may be used for an embedder component that processes feature data 250 corresponding to categorical features, where each category is mapped to a learned vector representation. In these and other embodiments, the embedder component that processes feature data 250 corresponding to numerical features may include fully connected neural network layers. These layers can project the numerical features into a higher-dimensional space.

[0102] Furthermore, the embedder component that processes the feature data 250 corresponding to features that are sequential or textual in nature may include specialized neural networks, and convolutional or recurrent neural networks may be utilized. These networks are adept at processing data given in sequences such as text or time series, and can capture the temporal dynamics or contextual nuances present in such data.

[0103] In some embodiments, the output from each embedder component may be a vector representation of its respective feature. In these and other embodiments, once individual feature embeddings are generated, they are combined to form a combined feature embedding 254. This combination can be achieved through methods such as concatenation or element-wise addition. Concatenation involves joining the embeddings end-to-end, which stores the individual feature representations within a larger embedding. Element-wise addition, on the other hand, merges the embeddings by adding up their corresponding elements, which can help highlight characteristics shared between features. The combined feature-event embedding 254 may represent a comprehensive encoding of a particular event.

[0104] In some embodiments, the combined feature event embeddings 254 may be further processed through a feedforward section 258 of the event representation module 204. The feedforward section 258 may include one or more feedforward layers capable of processing the combined feature event embeddings 254 and capturing interactions between features. For example, in some embodiments, the feedforward section 258 may include a feedforward neural network.

[0105] The feedforward section 258 handles combined feature-event embeddings by applying a set of transformations designed to capture the interactions between features. Each transformation within the feedforward section 258 may include a linear transformation followed by a nonlinear activation function. This allows the feedforward section 258 to learn complex patterns and dependencies between embedded features of a particular event.

[0106] In some embodiments, the feedforward section 258 may include multiple layers, each configured to progressively refine the feature interaction. The output of each layer may be passed to subsequent layers to allow for the construction of a deep representation of the interaction. This hierarchical processing helps capture both low-level and high-level feature interactions.

[0107] The output of the feedforward section 258 may be an event embedding 206, which may include a sophisticated set of feature embeddings representing the combined and interacted features of a particular event. In these and other embodiments, the event embedding 206 may be constructed as a fixed-dimensional vector representing the entire event, incorporating information from all of its constituent features.

[0108] This embedding process may allow a generative AI model (e.g., generative AI model 108) to effectively process a particular event and / or other events that may be represented in the partial process flow data 202. For example, event embedding 206 may capture semantic relationships between feature values ​​and allow the generative AI model to generalize through similar events.

[0109] Figure 2 shows a specific arrangement of components and data flows, but those skilled in the art will recognize that alternative configurations are possible to achieve similar functionality for predicting process flow completion and extracting relevant features. The core concept of generating event embeddings 206 can be implemented through various architectural designs without departing from the spirit or scope of this disclosure.

[0110] Figure 3 shows an exemplary system 300 related to generating one or more predictive event representations according to one or more embodiments of the present disclosure. In these and other embodiments, the system 300 may be related to the operations performed by the generating AI model 108 of Figure 1.

[0111] In some embodiments, the system 300 may include an autoregressive transformer decoder 308 ("decoder 308"), which may be an example of the generative AI model 108 in Figure 1. In these and other embodiments, the decoder 308 may be configured to process a partial event sequence 306 ("event sequence 306") to generate a predicted event embedding 310, which may be the exemplary predicted event representation 110 in Figure 1.

[0112] The event sequence 306 may include event embeddings of events that have already completed and / or have already been predicted. The event sequence 306 may indicate an ordered relationship between corresponding events. In these and other embodiments, the ordered nature of the event sequence 306 preserves the temporal relationships between events. Each event embedding may encode information about when that event occurred relative to other events in the sequence. Thus, this temporal ordering can provide context for the decoder 308 when predicting subsequent events.

[0113] In some examples, event sequence 306 may include event embeddings for a series of events that have already occurred in an ongoing process or incident (e.g., one or more event embeddings, such as event embedding 206 in Figure 2). Additionally or alternatively, event sequence 306 may include embeddings for events that were previously predicted but have not yet actually occurred (e.g., one or more previously determined predicted event embeddings, previously determined by decoder 308 in accordance with the operation described herein). For example, event sequence 306 may include embeddings up to event 1, event 2, and event p, where "p" represents the most recent event that has occurred and been logged, or the most recently predicted event. This allows encoder 308 to base its predictions on those predictions when predicting subsequent events in the sequence. In this disclosure, references to "event sequence 306" may also refer to any corresponding data (e.g., the events contained in event sequence 306, and the corresponding data, e.g., feature data corresponding to the events contained in event sequence 306). Furthermore, the term “event sequence 306” may be used to refer to the data corresponding to the event sequence after it has been processed by one or more intermediate components of the decoder 308.

[0114] The decoder 308 may process the event sequence 306 through a series of operations to generate a predicted event embedding 310. For example, in some embodiments, the decoder 308 may process the event sequence 306 using a masked multi-head self-focus layer 354 ("focus layer 354"). The focus layer 354 can help ensure that when the decoder 308 makes predictions, it uses only information from the events contained in the event sequence 306.

[0115] In these and other embodiments, after the focus layer 354, the decoder 308 may apply normalization layers 356 and 360 and a feedforward layer 358 to the event sequence 306. These layers may help transform the focus representation from the focus layer 354 and extract higher-level features. The decoder 308 may repeat this process multiple times, with a feedforward layer following self-focus, each iteration being referred to as a decoder iteration.

[0116] As event sequence data flows through decoder iterations, it may become increasingly contextualized based on the complete sequence history. The final iteration may generate a contextual representation of a predicted event that could be predicted as the next event in event sequence 306. Such a contextual representation may be projected into a specific embedding space (for example, as a specific vector type), which may be output as a predicted event embedding 310. The predicted event embedding 310 may represent a prediction for the next event in event sequence 306.

[0117] By processing the event sequence 306 in this autoregressive manner, the decoder 308 can effectively model the dependencies between events and generate predictions that take the complete contextual history into account. Masked attention and layer-wise processing allow the decoder 308 to construct increasingly sophisticated representations to make accurate predictions of the next event.

[0118] Figure 3 shows a specific arrangement of components and data flows, but those skilled in the art will recognize that alternative configurations are possible to achieve similar functionality for predicting process flow completion and extracting relevant features. The core concept of generating predictive event embeddings 310 can be implemented through various architectural designs without departing from the spirit or scope of this disclosure.

[0119] Figure 4 shows an exemplary system 400 relating to extracting one or more predicted features 416 from a predicted feature representation 410, according to one or more embodiments of the present disclosure. In these and other embodiments, the system 400 may relate to the operations performed by the feature extraction module 114 of Figure 1.

[0120] In some embodiments, the system 400 may include a feature extraction module 414, which may be an example of the feature extraction module 114 in Figure 1. In these and other embodiments, the feature extraction module 414 may be configured to process a particular predicted event representation 410 to extract one or more predicted features 416. For ease of explanation, the system 400 is described in relation to extracting predicted features 416 for a particular predicted event representation 410, but the described operation may be performed for any number of predicted event representations to generate any corresponding number of predicted features 416.

[0121] The predicted event representation 410 may be similar to the predicted event representation 110 in Figure 1 and / or the predicted event embedding 310 in Figure 3. Thus, the predicted event representation 410 can be organized as a vector that can represent different features of a particular event. Generally, the feature extraction module 414 may be configured to analyze such a vector and extract information corresponding to different features that may be encoded in the predicted event representation 410.

[0122] For example, in some embodiments, the feature extraction module 414 may receive the predicted event representation 410 as a high-dimensional vector encoding information about a specific predicted event.

[0123] The feature extraction module 414 can then apply a series of linear transformations to this vector representation. Each linear transformation may correspond to a different feature of interest, such as event type, timestamp, resource allocation, or other relevant attributes. These linear transformations can project the higher-dimensional representation onto a lower-dimensional space corresponding to the individual features.

[0124] After the linear transformation, the feature extraction module 414 may apply activation functions to the output. These activation functions can help capture nonlinear relationships in the data and map the output to an appropriate range for each feature type. For example, for categorical features such as event types, softmax activation may be used to generate a probability distribution across possible categories. Additionally or alternatively, for numerical features such as timestamps, different activations such as ReLU or sigmoid may be more appropriate.

[0125] The feature extraction module 414 may also incorporate additional processing steps tailored to specific feature types. For example, it may apply normalization techniques to ensure that numerical predictions fall within the expected range based on training data statistics. For categorical features, thresholding may be incorporated to generate discrete predictions from probability distributions.

[0126] The next event feature prediction component of the predicted feature 416 may include a variety of information about specific features that can be predicted for a particular predicted event corresponding to the predicted event representation 410. For example, different “next event feature predictions” may include predictions of event types or activities and provide probabilities of different possible activities that may occur as part of such an event. In these and other embodiments, such predictions may also include temporal information such as the expected timestamp of the next event or the duration until it occurs.

[0127] Additionally or alternatively, the following event feature predictions may include resource-related predictions that potentially indicate, for example, which teams, individuals, or systems are most likely to be involved in a particular predicted event. Feature predictions may also include estimates of various performance metrics or key performance indicators relevant to the specific process being modeled. Furthermore, the following event feature predictions may include predicted attributes such as the event's priority level, status indicator, or other domain-specific characteristics within a given process context.

[0128] In some implementations, feature predictions 416 may also include confidence scores or uncertainty estimates associated with each predicted next event feature. This can provide valuable information about the reliability of different aspects of the prediction, potentially allowing for more informed decision-making.

[0129] Figure 4 shows a specific arrangement of components and data flows, but those skilled in the art will recognize that alternative configurations are possible to achieve similar functionality for predicting process flow completion and extracting relevant features. The core concept of extracting predicted features 416 can be implemented through various architectural designs without departing from the spirit or scope of this disclosure.

[0130] Figure 5 shows an exemplary flowchart of an exemplary method 500 for generating a predicted sequence of events for a process flow, according to one or more embodiments of the present disclosure. One or more operations of method 500 can be implemented using any suitable system, apparatus, or device, such as those described with respect to Figures 1-4 and / or Figure 6. Although shown as discrete steps, the various steps of method 500 may be divided into additional steps, combined into fewer steps, or eliminated, depending on the desired implementation. Furthermore, the order in which the different steps are executed may vary depending on the desired implementation.

[0131] Method 500 may include block 502, in which a first representation of the first event of the process flow corresponding to the incident may be generated. This operation may correspond to the event representation module 104 in Figure 1, which generates an event representation 106 about the event from partial process flow data 102.

[0132] In block 504, the method may include using an artificial intelligence model to generate a second representation of a second event, which is a prediction of a future event following a first event in the process flow, based on a first representation. This operation may utilize the generative AI model 108 shown in Figure 1 to generate a predicted event representation 110 based on the event representation 106.

[0133] In block 506, the method may include using an artificial intelligence model to generate a third representation of a third event, which is a prediction of a future event following a second event in a process flow, based on a first and second representation. This operation may involve the generating AI model 108 predicting another predictive event representation 110 based on one or more previously generated predictive event representations 110 and event representations 106.

[0134] In block 508, the method may include generating a predicted event sequence for a process flow as a whole (e.g., the entire process flow) based on the predicted order of the first, second, and third events. This operation may correspond to generating the predicted complete process flow 120 shown in Figure 1.

[0135] In block 510, the method may include adjusting one or more allocations for the future use of one or more resources used to execute a process flow, based on a predicted event sequence. This operation may involve analyzing the predicted complete process flow 120 using system 112 in Figure 1 and adjusting resource allocations based on predicted features 116.

[0136] Method 500 may be modified, added to, or omitted without departing from the scope of this disclosure. For example, the steps and operations outlined are provided only as examples, and some of the steps and operations may be optionally combined into fewer steps and operations or expanded into additional steps and operations without impairing the essence of the disclosed embodiments.

[0137] For example, in some embodiments, method 500 may include additional actions that can be performed in connection with or as part of the actions shown in Figure 5.

[0138] For example, in some embodiments, method 500 may further include extracting feature data corresponding to the second event, based on the second representation, which includes information corresponding to a plurality of second features corresponding to the second event. This operation may be performed by the feature extraction module 114 shown in Figure 1. The extracted feature data may include various attributes of the predicted second event, such as activity type, resource allocation, timing information, or other relevant characteristics.

[0139] In these and other embodiments, at least one of one or more resources for which one or more allocations need to be adjusted may be identified based on second feature data. This operation may use the extracted feature data to determine which particular resources may require allocation adjustment based on the predicted characteristics of the second event.

[0140] Additionally or alternatively, method 500 may further include obtaining first feature data corresponding to a first event, the first feature data containing first information corresponding to a plurality of first features corresponding to the first event. This operation may involve processing the partial process flow data 102 shown in Figure 1 to extract relevant features from the first event.

[0141] In these and other embodiments, the first representation may be generated based on the first feature data. This operation can utilize the event representation module 104 shown in Figure 1 to create an embedding or other suitable representation of the first event based on its extracted features.

[0142] Additionally or alternatively, Method 500 may further include generating a first representation based on first feature data by generating individual embeddings for each of the multiple first features. This operation may correspond to the individual embedder components shown in layer 252 of Figure 2, where each feature is processed separately. In these and other embodiments, Method 500 may include combining the individual embeddings into the first representation, as described with respect to Figure 2.

[0143] In these and other embodiments, the artificial intelligence model used in Method 500 may be trained on features extracted from multiple completed events, each corresponding to a multiple completed incident. This training data may be derived from historical process logs of fully completed incidents, for example, as described with respect to the training of the generating AI model 108 in Figure 1.

[0144] Additionally or alternatively, training may be based on an event sequence of completed events corresponding to individual completed incidents. This training data may include a complete sequence of events from start to finish for multiple historical incidents, as described, for example, in relation to training the generating AI model 108 in Figure 1.

[0145] Figure 6 shows a block diagram of an exemplary computing system 602 according to at least one embodiment of the present disclosure. The computing system 602 may be configured to perform or direct one or more appropriate operations described in the present disclosure. In some embodiments, the computing system 602 may be used to control operations related to generating training data (e.g., labels and / or features for training one or more machine learning models). The computing system 602 may include a processor 650, memory 652, and data storage 654. The processor 650, memory 652, and data storage 654 may be communicatively coupled.

[0146] Generally, the processor 650 may include any suitable computer, computing entity, or processing device, including various computer hardware or software modules, and may be configured to execute instructions stored in any applicable computer-readable storage medium. For example, the processor 650 may include a microprocessor, microcontroller, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or any other digital or analog circuitry configured to interpret and / or execute program instructions and / or process data. Although shown as a single processor in Figure 6, the processor 650 may include any number of processors configured to individually or collectively perform or instruct any number of operations described in this disclosure. In addition, one or more of the processors may reside on one or more different electronic devices, such as different servers.

[0147] In some embodiments, the processor 650 may be configured to interpret and / or execute program instructions, and / or process data stored in memory 652 or data storage 654 or memory 652 and data storage 654. In some embodiments, the processor 650 may fetch program instructions from data storage 654 and load the program instructions into memory 652. After the program instructions are loaded into memory 652, the processor 650 may execute the program instructions.

[0148] Memory 652 and data storage 654 may include computer-readable storage media that carry or store computer-executable instructions or data structures. Such computer-readable storage media may include, but are not limited to, random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM®), compact disk read-only memory (CD-ROM) or other optical disk storage, magnetic disk storage or other magnetic storage devices, flash memory devices (e.g., solid-state memory devices), or any other non-temporary storage media that may be used to store specific program code in the form of computer-executable instructions or data structures and that can be accessed by a general-purpose or dedicated computer. In these and other embodiments, the term “non-temporary” as described herein should be interpreted to exclude only the types of temporary media that were found to be outside the scope of patentable subject matter in the Federal Circuit Court of Appeals decision In re Nuijten, 500 F.3d 1346 (Fed. Cir. 2007).

[0149] The above combinations may also fall within the scope of computer-readable storage media. Computer-executable instructions may include, for example, instructions and data configured to execute an operation or group of operations on the processor 650.

[0150] Without departing from the scope of this disclosure, modifications, additions, or omissions may be made to, for example, the computing system 602. For example, in some embodiments, the computing system 602 may include any number of other components that are not expressly illustrated or described.

[0151] The terms used in this disclosure, and in particular in the attached claims (e.g., the body of the attached claims), are generally intended to be “open” terms (for example, “contains” should be interpreted as “contains, but is not limited to…”, “has” should be interpreted as “has at least…”, and “contains” should be interpreted as “contains, but is not limited to…”).

[0152] Furthermore, if a specific number of claims to be introduced is intended, such intention is explicitly stated in the claim; if no such statement is present, such intention does not exist. For example, to aid understanding, the following appended claims may include the use of the introductory phrases “at least one” and “one or more” to introduce a claim. However, the use of such phrases should not be interpreted as implying that the introduction of a claim by the indefinite article “a” or “an” limits any particular claim containing such introduced claim to embodiments containing only one such claim, even if the same claim contains the introductory phrase “one or more” or “at least one” and an indefinite article such as “a” or “an” (for example, “a” and / or “an” should be interpreted as meaning “at least one” or “one or more”). The same applies to the use of the definite article used to introduce a claim.

[0153] Furthermore, even if a specific number of claims to be introduced is explicitly stated, a person skilled in the art will recognize that such a statement should be interpreted as meaning at least that number (for example, the statement “two statements” without other modifiers means at least two statements, or two or more statements). Furthermore, when idiomatic expressions similar to “at least one of A, B, and C, etc.” or “one or more of A, B, and C, etc.” are used, such constructions are generally intended to include A only, B only, C only, A and B together, A and C together, B and C together, or A, B and C together, etc.

[0154] Furthermore, any disjunctive phrase presenting two or more alternative terms should be understood to mean that in this paper, the claims, or the drawings, it is possible to include one of the terms, either one of the terms, or both of the terms. For example, the phrase "A or B" should be understood to mean that it includes the possibilities of "A" or "B" or "A and B," even if the terms "and / or" are used elsewhere.

[0155] All examples and conditional statements contained herein are intended for educational purposes to help readers understand the concepts described herein and the inventions that contribute to the advancement of the art, and should be construed as not being limited to such specifically described examples and conditions. While embodiments of this disclosure have been described in detail, various changes, substitutions, and modifications can be made without departing from the spirit and scope of this disclosure.

[0156] With regard to embodiments including the above examples, the following additional information is disclosed. (Note 1) The first stage of the process flow for responding to an incident is to generate the first representation of the first event; A step of using an artificial intelligence model to generate a second representation of a second event, which is a prediction of future events following the first event in the process flow, based on the first representation; A step of generating a third representation of a third event, which is a prediction of a future event following the second event in the process flow, based on the first and second representations and using the artificial intelligence model; A step of generating a predicted event sequence for the entire process flow based on the predicted order of the first event, the second event, and the third event; A step of adjusting one or more allocations for the future use of one or more resources used to execute the process flow, based on the predicted event sequence. Methods that include... (Note 2) Based on the second representation, the step of extracting feature data that corresponds to the second event and includes information corresponding to a plurality of second features corresponding to the second event; A step of identifying at least one of the one or more resources for the step of adjusting the one or more allocations based on the characteristic data. The method described in Appendix 1, further including the method described in Appendix 1. (Note 3) The aforementioned first event is the method described in Appendix 1, which has already been completed. (Note 4) The method described in Appendix 1, wherein the first event is a future event that has not yet occurred. (Note 5) The method according to Appendix 1, further comprising the step of obtaining first feature data corresponding to the first event, wherein the first feature data includes first information corresponding to a plurality of first features corresponding to the first event, and the first representation is generated based on the first feature data. (Note 6) The first representation is based on the first feature data and comprises at least: For each of the aforementioned plurality of first features, an individual embedding is generated; By combining the individual embeddings in the first representation The method described in Appendix 5 is used to generate the generated product. (Note 7) The aforementioned artificial intelligence model: Features extracted from multiple completed events, each corresponding to a different completed incident; or Event sequence of the completed events corresponding to each completed incident The method described in Appendix 1, which is trained based on one or more of the following. (Note 8) One or more non-temporary computer-readable media storing instructions that cause a system to perform an action in response to being executed by one or more processors, wherein the action is: The first stage of the process flow for responding to an incident is to generate the first representation of the first event; A step of using an artificial intelligence model to generate a second representation of a second event, which is a prediction of future events following the first event in the process flow, based on the first representation; A step of generating a third representation of a third event, which is a prediction of a future event following the second event in the process flow, based on the first and second representations and using the artificial intelligence model; A step of generating a predicted event sequence for the entire process flow based on the predicted order of the first event, the second event, and the third event; A step of adjusting one or more allocations for the future use of one or more resources used to execute the process flow, based on the predicted event sequence. Non-temporary computer-readable media, including [specific examples of such media]. (Note 9) The aforementioned operation is: Based on the second representation, the step of extracting feature data that corresponds to the second event and includes information corresponding to a plurality of second features corresponding to the second event; A step of identifying at least one of the one or more resources for the step of adjusting the one or more allocations based on the characteristic data. Non-temporary computer-readable media as described in Appendix 8, including, furthermore. (Note 10) The aforementioned first event is a non-temporary computer-readable medium as described in Appendix 8, which has already been completed. (Note 11) The aforementioned first event is a future event that has not yet occurred, as described in Appendix 8, and is a non-transient computer-readable medium. (Note 12) The operation further includes the step of acquiring first feature data corresponding to the first event, wherein the first feature data includes first information corresponding to a plurality of first features corresponding to the first event, and the first representation is generated based on the first feature data, as described in Appendix 8, for a non-temporary computer-readable medium. (Note 13) The first representation is based on the first feature data and comprises at least: For each of the aforementioned plurality of first features, an individual embedding is generated; By combining the individual embeddings in the first representation The non-temporary computer-readable media generated as described in Appendix 12. (Note 14) The aforementioned artificial intelligence model: Features extracted from multiple completed events, each corresponding to a different completed incident; or Event sequence of the completed events corresponding to each completed incident A non-temporary computer-readable medium as described in Appendix 8, which is trained on one or more of the following. (Note 15) One or more processors; and Having one or more computer-readable media that store computer-executable instructions causing the one or more processors to perform an operation in response to being executed by the one or more processors. A system, wherein the operation is: The first stage of the process flow for responding to an incident is to generate the first representation of the first event; A step of using an artificial intelligence model to generate a second representation of a second event, which is a prediction of future events following the first event in the process flow, based on the first representation; A step of generating a third representation of a third event, which is a prediction of a future event following the second event in the process flow, based on the first and second representations and using the artificial intelligence model; A step of generating a predicted event sequence for the entire process flow based on the predicted order of the first event, the second event, and the third event; A step of adjusting one or more allocations for the future use of one or more resources used to execute the process flow, based on the predicted event sequence. A system that includes this. (Note 16) The aforementioned operation is: Based on the second representation, the step of extracting feature data that corresponds to the second event and includes information corresponding to a plurality of second features corresponding to the second event; A step of identifying at least one of the one or more resources for the step of adjusting the one or more allocations based on the characteristic data. The system described in Appendix 15, which further includes the system described in Appendix 15. (Note 17) The aforementioned first event is the system described in Appendix 15, which has already been completed. (Note 18) The aforementioned first event is a future event that has not yet occurred, as described in Appendix 15 of the system. (Note 19) The system as described in Appendix 15, wherein the operation further includes the step of acquiring first feature data corresponding to the first event, the first feature data comprising first information corresponding to a plurality of first features corresponding to the first event, and the first representation being generated based on the first feature data. (Note 20) The first representation is based on the first feature data and comprises at least: For each of the aforementioned plurality of first features, an individual embedding is generated; By combining the individual embeddings in the first representation The system generated is described in Appendix 19. [Explanation of Symbols]

[0157] 102 Partial Process Flow Data 104 Event Representation Module 106 Event Representation 108 Generative AI Models 110 Predictive Event Representation 112 System 114 Feature Extraction Models 116 Predicted Features 120 Predicted Complete Process Flow 202 Partial Process Flow Data 206 Event Embedding 250 Feature Data / Event Features 252 Embedder 256 Event Data 258 Feedforward 310 Predicted Event Embedding 306 Events 354 Masked Multi-Head Attention 358 Feedforward 356 In addition, normalization 360 In addition, normalize 410 Predictive Event Representation 414 Linear 416 predicted features Generate the first representation of the first event in the process flow for responding to a 502 incident. 504 Using an artificial intelligence model, generate a second representation of the second event, which is a prediction of future events following the first event in the process flow, based on the first representation. 506 Using an artificial intelligence model, generate a third representation of a third event, which is a prediction of future events following the second event in the process flow, based on the first and second representations. 508 As predicted, generate a predicted event sequence for the entire process flow based on the order of the first, second, and third events. 510 Based on the predicted event sequence, adjust one or more allocations for the future use of one or more resources used to execute the process flow. 602 Computing Systems 650 Processor 652 memory 654 Data Storage

Claims

1. The first step in the process flow for responding to an incident is to generate the first representation of the first event; A step of using an artificial intelligence model to generate a second representation of a second event, which is a prediction of future events following the first event in the process flow, based on the first representation; A step of generating a third representation of a third event, which is a prediction of a future event following the second event in the process flow, based on the first and second representations and using the artificial intelligence model; A step of generating a predicted event sequence for the entire process flow based on the predicted order of the first event, the second event, and the third event; A step of adjusting one or more allocations for the future use of one or more resources used to execute the process flow, based on the predicted event sequence. Methods that include...

2. Based on the second representation, the step of extracting feature data that corresponds to the second event and includes information corresponding to a plurality of second features corresponding to the second event; A step of identifying at least one of the one or more resources for the step of adjusting the one or more allocations based on the characteristic data. The method according to claim 1, further comprising:

3. The method according to claim 1, wherein the first event has already been completed.

4. The method according to claim 1, wherein the first event is a future event that has not yet occurred.

5. The method according to claim 1, further comprising the step of obtaining first feature data corresponding to the first event, wherein the first feature data includes first information corresponding to a plurality of first features corresponding to the first event, and the first representation is generated based on the first feature data.

6. The first representation is based on the first feature data and comprises at least: For each of the aforementioned plurality of first features, an individual embedding is generated; By combining the individual embeddings in the first representation The method according to claim 5, which is generated.

7. The aforementioned artificial intelligence model is: Features extracted from multiple completed events, each corresponding to a separate completed incident; or Event sequence of the completed events corresponding to each completed incident The method according to claim 1, which is trained based on one or more of the following.

8. One or more non-temporary computer-readable media storing instructions that cause a system to perform an action in response to being executed by one or more processors, wherein the action is: The first step in the process flow for responding to an incident is to generate the first representation of the first event; A step of using an artificial intelligence model to generate a second representation of a second event, which is a prediction of future events following the first event in the process flow, based on the first representation; A step of generating a third representation of a third event, which is a prediction of a future event following the second event in the process flow, based on the first and second representations and using the artificial intelligence model; A step of generating a predicted event sequence for the entire process flow based on the predicted order of the first event, the second event, and the third event; A step of adjusting one or more allocations for the future use of one or more resources used to execute the process flow, based on the predicted event sequence. Non-temporary computer-readable media, including [specific examples of such media].

9. The aforementioned operation is: Based on the second representation, the step of extracting feature data that corresponds to the second event and includes information corresponding to a plurality of second features corresponding to the second event; A step of identifying at least one of the one or more resources for the step of adjusting the one or more allocations based on the characteristic data. A non-temporary computer-readable medium according to claim 8, further comprising:

10. The non-temporary computer-readable medium according to claim 8, wherein the first event has already been completed.

11. The non-temporary computer-readable medium according to claim 8, wherein the first event is a future event that has not yet occurred.

12. The non-temporary computer-readable medium according to claim 8, wherein the operation further includes the step of acquiring first feature data corresponding to the first event, the first feature data comprising first information corresponding to a plurality of first features corresponding to the first event, and the first representation is generated based on the first feature data.

13. The first representation is based on the first feature data and comprises at least: For each of the aforementioned plurality of first features, an individual embedding is generated; By combining the individual embeddings in the first representation A non-temporary computer-readable medium generated according to claim 12.

14. The aforementioned artificial intelligence model is: Features extracted from multiple completed events, each corresponding to a separate completed incident; or Event sequence of the completed events corresponding to each completed incident A non-temporary computer-readable medium according to claim 8, which is trained on one or more of the following.

15. One or more processors; and Having one or more computer-readable media that store computer-executable instructions causing the one or more processors to perform an operation in response to being executed by the one or more processors. A system, wherein the operation is: The first step in the process flow for responding to an incident is to generate the first representation of the first event; A step of using an artificial intelligence model to generate a second representation of a second event, which is a prediction of future events following the first event in the process flow, based on the first representation; A step of generating a third representation of a third event, which is a prediction of a future event following the second event in the process flow, based on the first and second representations and using the artificial intelligence model; A step of generating a predicted event sequence for the entire process flow based on the predicted order of the first event, the second event, and the third event; A step of adjusting one or more allocations for the future use of one or more resources used to execute the process flow, based on the predicted event sequence. A system that includes this.

16. The aforementioned operation is: Based on the second representation, the step of extracting feature data that corresponds to the second event and includes information corresponding to a plurality of second features corresponding to the second event; A step of identifying at least one of the one or more resources for the step of adjusting the one or more allocations based on the characteristic data. The system according to claim 15, further comprising:

17. The system according to claim 15, wherein the first event has already been completed.

18. The system according to claim 15, wherein the first event is a future event that has not yet occurred.

19. The system according to claim 15, wherein the operation further includes the step of acquiring first feature data corresponding to the first event, the first feature data comprising first information corresponding to a plurality of first features corresponding to the first event, and the first representation is generated based on the first feature data.

20. The first representation is based on the first feature data and comprises at least: For each of the aforementioned plurality of first features, an individual embedding is generated; By combining the individual embeddings in the first representation The system according to claim 19, which is generated.

Images