Abstract:Existing well investigated Predictive Process Monitoring techniques typically construct a predictive model based on past process executions, and then use it to predict the future of new ongoing cases, without the possibility of updating it with new cases when they complete their execution. This can make Predictive Process Monitoring too rigid to deal with the variability of processes working in real environments that continuously evolve and/or exhibit new variant behaviours over time. As a solution to this problem, we evaluate the use of three different strategies that allow the periodic rediscovery or incremental construction of the predictive model so as to exploit new available data. The evaluation focuses on the performance of the new learned predictive models, in terms of accuracy and time, against the original one, and uses a number of real and synthetic datasets with and without explicit Concept Drift. The results provide an evidence of the potential of incremental learning algorithms for predicting process monitoring in real environments.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: Existing Predictive Process Monitoring (PPM) models cannot be automatically updated to adapt to new data after being constructed, resulting in these models being too rigid when dealing with constantly changing and evolving real - world business processes. Specifically: 1. **Existing problems**: - Existing PPM techniques usually build prediction models based on past process execution data and use this model to predict the future situation (such as the result, remaining time or the next activity) of new ongoing cases. However, once the model is constructed, it does not automatically take into account the data of newly completed cases. - Such static models are not flexible enough in the face of constantly changing and evolving processes in the real environment because business processes in reality are complex, variable and lack a stable state. 2. **Proposed problems**: - How can predictive process monitoring models be made to dynamically adapt to new process execution data, so as to better cope with the constantly changing business processes in the real environment? - Specifically, the author explores how to maintain the currency and effectiveness of prediction models through different strategies (such as making no update, retraining without optimizing hyper - parameters, full retraining and incremental update), especially in the case of concept drift. 3. **Research objectives**: - Evaluate the impact of different update strategies on prediction quality, especially on data sets with and without concept drift. - Compare the differences in the time required for different update strategies to construct and update prediction models. Through these studies, the author hopes to find an update strategy that can efficiently adapt to new data while ensuring prediction quality, thereby improving the flexibility and accuracy of predictive process monitoring models in practical applications.

How do I update my model? On the resilience of Predictive Process Monitoring models to change

A novel lifelong machine learning-based method to eliminate calibration drift in clinical prediction models.

Is Your Anomaly Detector Ready for Change? Adapting AIOps Solutions to the Real World

Enhancing Model Adaptability Using Concept Drift Detection for Short-Term Load Forecast

Predictive Process Model Monitoring using Recurrent Neural Networks

Towards cost-sensitive adaptation: When is it worth updating your predictive model?

Measuring the Stability of Process Outcome Predictions in Online Settings

Online Updating Soft Sensor Modeling and Industrial Application Based on Selectively Integrated Moving Window Approach

Robust Dynamic Latent Variable Model for Dynamic Process Monitoring with Missing Data in Cyber-Physical Systems

Adaptive Monitoring Method for Batch Processes Based on Phase Dissimilarity Updating with Limited Modeling Data

Recursive Gaussian Mixture Models for Adaptive Process Monitoring

How to Sustainably Monitor ML-Enabled Systems? Accuracy and Energy Efficiency Tradeoffs in Concept Drift Detection

Remaining Useful Life Prediction for a Nonlinear Heterogeneous Wiener Process Model With an Adaptive Drift

On the Model Update Strategies for Supervised Learning in AIOps Solutions

Automating concept-drift detection by self-evaluating predictive model degradation

Adaptive Nonlinear Model Predictive Control Using an On-line Support Vector Regression Updating Strategy

Model updating after interventions paradoxically introduces bias

Predictive process model monitoring using long short-term memory networks

Data-Driven Model Predictive Monitoring for Dynamic Processes

An LSTM-based predictive monitoring method for data with time-varying variability

Learning-Based Approaches to Predictive Monitoring with Conformal Statistical Guarantees