Prediction-based Resource Allocation using LSTM and Minimum Cost and Maximum Flow Algorithm

Real-life business process event logs may suffer from significant data quality problems negatively influencing process mining analysis. Over time, a range of approaches has been developed to detect and repair these quality problems. Validation of these approaches tends to be challenging due to the lack of a ground truth. Moreover, the identification and definition of event log quality problems have been tackled mainly through a pattern-based approach, with systematic and extensible methods currently lacking. In this article, we present FLAWD , a formal language for describing event log data quality issues that enables solutions addressing the shortcomings of process mining data quality research identified above. FLAWD can be used to formally describe and possibly reason over event log data quality errors, as well as to guide the development of tools for controlled and sophisticated “polluting” of event logs through which benchmark datasets may be systematically created. We present the abstract syntax grammar of FLAWD and an open-source software tool based on it that allows for the insertion of all so-called event log imperfection patterns in a stochastic manner. We show how FLAWD has been used in our research to generate benchmark datasets and how it can be used to formally describe and replicate a range of errors found in real-life event logs.

In this paper, we focus on the dynamic traffic assignment (DTA) problem on large-scale expressway networks especially under the condition of traffic events (such as severe weather, large traffic accidents etc.). We formulate the expressway network DTA problem as a nonconvex optimization problem. Considering the difficulty of solving the problem, we put forward an approximate solution algorithm based on the network flow theory with high computational efficiency and strong applicability. Specifically, a machine learning approach and a series of customization strategies of network flow model (including multi-origin and multidestination split, flow discretization and parallel edges addition methods) are proposed to transform the DTA problem into a resoluble network flow problem. After that, in order to improve the operational efficiency of an expressway network after traffic events occurring, the proposed network flow based two-layer DTA model (NTDM) is separated into two parts-traffic flow limitation model based on maximum flow algorithm and traffic assignment model based on minimum cost flow algorithm. Experiments on real expressway network data in some provinces of China show that our method can make a significant 26.03% reduction in saturation rate in a particular scenario. In efficiency test, the NTDM is at least 600 times faster than the traditional analytic method in a network with thousands of nodes.

Workflow management is an important technology of business process management that links tasks and qualified resources as a bridge. Researches have been carried out to improve the resource allocation of workflow that is often performed manually and empirically either by mining resource allocation rules or by optimizing the resource allocation for tasks to achieve certain goals such as minimal cost or duration. None of these approaches can guarantee to give the suitable solution to resource allocators because of the dynamic natures of business process executions. In this paper we propose an approach, BNRR (Bayesian Network-based Resource Recommendation), to recommend the most proficient sets of resources for a business process based on event logs, which gives the allocators chances to find the most suitable solution. Our approach considers both the information about the resource dependency and the information about the resource capability. The approach can be applied to recommend resources either for a whole workflow or for an individual task. The approach is validated by experiments on real life data.KeywordsResource recommendationstaff assignmentBayesian Networkbusiness process management

The TanDEM-X Mission has as primary objective to generate a high resolution global Digital Elevation Model. This paper proposes a new method for multibaseline Phase Unwrapping which is the critical point of this generation. We propose to combine both Minimum Cost Flow (MCF) and Maximum a Posteriori (MAP) estimation. The latter is used to solve phase gradient ambiguities. The problem is posed as an energy minimization one and solved using Belief Propagation (BP) which is an iterative process. Nevertheless, although very good results are obtained on loopy graphs, it is not guaranteed to converge. Thus, phase unwrapping of the most accurate interferogram is finally performed with the MCF algorithm and takes as input the unwrapped gradients.

A theoretical and experimental study of a new algorithm for minimum cost flow in dynamic graphs

Flexible business processes can often be modelled more easily using a declarative rather than a procedural modelling approach. Process mining aims at automating the discovery of business process models. Existing declarative process mining approaches either suffer from performance issues with real-life event logs or limit their expressiveness to a specific set of constaint types. Lately, RelationalXES, a relational database architecture for storing event log data, has been introduced. In this paper, we introduce a mining approach that directly works on relational event data by querying the log with conventional SQL. By leveraging database performance technology, the mining procedure is fast without limiting itself to detecting certain control-flow constraints. Queries can be customised and cover process perspectives beyond control flow, e.g., organisational aspects. We evaluated the performance and the capabilities of our approach with regard to several real-life event logs.

The goal of process mining is to discover the process behavior from the runtime information of process executions. Having labeled process logs is crucial for process mining research. However, real life event logs at process-aware information systems are mostly partially assigned to case identifiers, known as unlabeled event log problem. As a remedy to labeled data need in process mining research, we propose an approach to generate synthetic event logs according to the provided process profile, which outlines the activity vocabulary and structure of the corresponding business process. We evaluate the performance of our prototypical implementation in term of compatible log generation under varying parameter setting complexities.

Process mining, which aims to mine a high-quality process model from event log, provides a powerful tool to support the design, enactment, management, and analysis of operational business processes. However, the task is not easy because the algorithm needs to discover various complex process structures, handle noisy and incomplete event logs and balance multiple performance indicators. In this paper, a novel algorithm (called PSOMiner) for process mining is proposed, which consists of a discrete particle swarm optimization algorithm and guided local mutation. The former is in charge of searching the solution space of causal matrix and the latter is used to help the algorithm skip out the local optimum when it suffers from premature. A fine-grained scoring strategy which used to assign a score to each position of a particle (i.e. causal matrix) is presented to direct the mutation. The experiments were performed on 28 synthetic event logs with/without noise and 4 real-life event logs, and three classical algorithms of process mining (ETM, Hybrid ILP Miner, HM) were chosen for comparison. The results show that (1) PSOMiner achieved the best f-score on 25 synthetic event logs; (2) The average f-score of PSOMiner is 0.825 on 4 real-life event logs, which is superior to ETM whose average f-score is 0.703.

With the development of data mining and machine learning, we can get much useful information from historical data. For a business process system, it maintains large amount of process execution data, especially records of events corresponding to the execution of activities, which can also be called event log. Predictive business process monitoring methods exploit logs of completed cases of a process in order to make predictions and recommendation about current running cases. This paper proposes an improved approach for process outcome prediction and next activity recommendation. It estimates the accuracy that a given goal will be fulfilled upon completion of a current running process case through three different methods. Each method includes both clustering phase and classification phase. However, different levels of historical data (business level and control flow level) in event log are used, and the size of data and number of features also differs. We show our improved approach to deal with historical log, encode each feature vector, train predictive model and how to use trained models for predicting the outcome of current case and recommending the next event. Finally, through a series of experiment, we compare three different method and existing approach.

Outcome-based predictive process monitoring deals with predicting the outcomes of running cases in a business process using feature vectors extracted from completed traces in an event log. Traditionally, in outcome-based predictive monitoring, a different model is developed using a bucket containing different types of feature vectors. This allows us to extend the traditional evaluation of the quality of process outcome predictions models beyond simply measuring the overall performance, developing a quality assessment framework based on three metrics: one considering the overall performance on all feature vectors, one considering the different levels of performance achieved on feature vectors belonging to individual buckets, i.e., the stability of the performance across buckets, and one considering the stability of the individual predictions obtained, accounting for how close the predicted probabilities are to the cutoff thresholds used to determine the predicted labels. The proposed metrics allow to evaluate, given a set of alternative designs, i.e., combinations of classifier and bucketing method, the quality of the predictions of each alternative. For this evaluation, we suggest using either the concept of Pareto-optimality or a scenario-based scoring method. We discuss an evaluation of the proposed framework conducted with real-life event logs.

Everything at the proper time: Repairing identical timestamp errors in event logs with Generative Adversarial Networks

In last few years lot of work is carried out in process mining field by various researchers. Process mining deals with analysis and extraction of process related information from the event logs created by business processes. Predictive monitoring of business process is subfield of process mining which includes activities where event logs are analyzed to make various process specific predictions. The various machine learning and deep learning techniques have been proposed in predictive business process monitoring (BPM). The aim of these techniques is to predict next process event, remaining cycle time, deadline violations etc. of running process instance. The goal of this paper is to discuss the most representative deep learning approaches used for the runtime prediction of business process. The different types of deep learning approaches used in predictive BPM based on Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM) and Stacked Autoencoders have are highlighted in this paper. Also we are focusing on aspects like type of dataset, predicted values, type of data encoding and quality evaluation metrics for the categorization of these approaches. In this paper we have highlighted various research gaps in mentioned deep learning approaches which can be referred by other researchers in this field to enhance effectiveness of predictive BPM.

We present an algorithm that computes exact maximum flows and minimum-cost flows on directed graphs with m edges and polynomially bounded integral demands, costs, and capacities in m 1+ o (1) time. Our algorithm builds the flow through a sequence of m 1+ o (1) approximate undirected minimum-ratio cycles, each of which is computed and processed in amortized m o (1) time using a new dynamic graph data structure. Our framework extends to algorithms running in m 1+ o (1) time for computing flows that minimize general edge-separable convex functions to high accuracy. This gives almost-linear time algorithms for several problems including entropy-regularized optimal transport, matrix scaling, p -norm flows, and p -norm isotonic regression on arbitrary directed acyclic graphs.

Goal modeling is widely used to align stakeholders requirements with architectural models. In contrast to goal models which are usually defined by stakeholders, architectural models are not always defined explicitly, or systems may not be used as they were designed. To understand the real behavior of a system, process mining techniques can be applied. These techniques allow us to automatically construct real models from the system's event logs. These real models can be compared to goal models to reveal their common and different parts. Unfortunately, graph-based approaches used for the model comparison are computationally expensive and can hardly be applied to large process models constructed from real-life event logs. In this paper, we adapt well-known heuristic approaches to efficiently compare process models. These approaches were tested on real data. The obtained results can be further used to compare real and expected (goal) process behavior.

A transshipment problem with demands that exceed network capacity can be solved by sending flow in several waves. How can this be done in the minimum number, T, of waves, and at minimum cost, if costs are piecewise linear convex functions of the flow? In this paper, we show that this problem can be solved using $\min\{ m,\log T,\ub{\Gamma}{U} \}$ maximum flow computations and one minimum (convex) cost flow computation. Here m is the number of arcs, $\Gamma$ is the maximum supply or demand, and U is the maximum capacity. When there is only one sink, this problem can be solved in the same asymptotic time as one minimum (convex) cost flow computation. This improves upon the previous best algorithm to solve the problem without costs by a factor of k. Our solutions start with a stationary fractional flow and use rounding to transform this into an integral flow. The rounding procedure takes O(n) time.