Horia Ciocarlie

Anomaly detection in data mining. Hybrid approach between filtering-and-refinement and DBSCAN

Anomaly detection is a domain that represents the key for the future of data mining. We will try to present some key anomaly detection methods applicable in the data mining process. Some methods are existing techniques as the DBSCAN algorithm and some have just been presented to the public recently and could be the answer to future anomaly detection development. One example is the filtering-and-refinement approach, a new general two stage technique for more efficient and effective anomaly detection. This paper will try to illustrate the strengths and weaknesses of the classical techniques presented but as we will see the results are completely dependent on the data sets that are analyzed. We will emphasize on efficiency, robustness and accuracy. We will also try to demonstrate a hybrid approach obtained by combining the filtering-and-refinement method with the DBSCAN algorithm. In our experiments we pursued to compare the performance of the normal DBSCAN algorithm with the performance of the hybrid one. Our results indicate that the hybrid method is more accurate in terms of detecting anomalies and far superior in terms of speed than the normal DBSCAN algorithm.

Merging business processes for a common workflow in an organizational collaborative scenario

In the current paper we propose an algorithm for merging business process models when the process similarity rate is high. We consider the representation of business processes in Business Process Modeling Notation format and we present a possible reduction of the elements to directed graph format components. A common graph extracted is post processed by applying a transitive reduction mechanism to obtain the simplest graph with the fewest possible edges that has the same reachability as the original graph. We present briefly other directions for merging business processes and we compare them with the proposed one. We start from the assumption that collaborations in between companies when the compatibility level from process perspective is high have better chances to succeed. Studies have shown that inter organizational collaborations fail mostly because of incompatibilities in methodology, lacks in transition process and failures in defining feasible targets. With common processes resulted from merging internally used process models, the transition setup is simple and effortless. The common process obtained could be manually or automatically updated for ensuring the compliance with shared and individual objectives and internal constraints.

Applied process mining in software development

Current study uses the event process logs of a custom software development process used internally in the IT department of a large automotive company. The target is to extract hidden information about the process, difficult to identify due to the wideness and complexity of the input data. The log file contains all incident records tracked during one release of a software product. This analysis was performed using process mining techniques and tools included in ProM Framework, an academic project of the Eindhoven Technical University. The paper describes the steps followed to extract the working process model, organizational network and statistical information. Based on the obtained results, an action plan is put in practice with the scope of improving the development process and the release result.

G-code optimization algorithm and its application on Printed Circuit Board drilling

Computer Numerical Controlled (CNC) machines are more and more used for a variety of tasks such as milling, drilling, lathing, etc. As many CNCs are programmed using G-Code, it is of great interest for this code to be close to the optimum, taking into consideration different metrics such as the total path length covered by the CNC head during the machining process or the total machining time. We propose an algorithm to optimize a large class of G-Codes, by reducing the total path length. As most of the machining operations require translations, our algorithm is applicable for many processes. In this paper we describe the algorithm and demonstrate its performance in the specific case of Printed Circuit Boards (PCB) drilling.

Similarity of business process models in a modular design

Using sub-processes in business process design enables flexibility, ensures a better understanding of the business scope and simplifies the complexity. A modular process design has several advantages in the area of scalability and re-usage of components. In the current paper we analyze the similarity of business processes when a modular design approach is used for process definition. We consider that the similarity of business processes is an important factor for the identification of the compatibility level between organizations. A better compatibility simplifies the setup in case of collaborations with shared resources and common business results. Process mining algorithms for process discovery and graph comparison techniques adapted are used for this purpose. Process definitions are extracted and reduced to graph format. Graphs are compared considering sub-graph composition and a business process similarity factor is extracted. For graph node match, semantic string similarity algorithms are evaluated and used further. This choice enables flexibility in comparing business processes from different business areas. The theoretical aspects are illustrated on a case study.

Maximizing the captured energy of a wind system

This paper shows the original methods to estimate the maximum energy of a wind system that operate at variable wind speed. The fundamental problem in any control system is the knowledge of rotation speed value for that the captured energy is maximal.

Business process similarity metric supporting one-to-many relationship

In many areas graph match techniques are used to compare and identify common characteristics. In this paper we apply graph similarity techniques on the business processes used inside organizations and extracted with process mining techniques. The scope is to identify if an organization uses a similar process for a specific business case as another organization. However as the existence of exact matching is less probable, error tolerant graph matching techniques are more suitable for real life data. Business processes could have a different granularity level; one business process is more detailed in specific areas than the business process subject of the comparison. The custom algorithm for business process match presented in this paper takes into consideration a one-to-many relation for activities: one activity is matched with a set of activities in the other graph. Such information is important in extracting the common characteristics of organizations and could represent an input for choosing a collaborator. Business processes if not available are extracted with process mining techniques and are reduced to directed graph format. A custom graph similarity algorithm extended for multivalent nodes is applied and a business process similarity factor is retrieved.

Complex network interpretation of European Union economic dynamics

The multitude of events that rapidly take place on the political scene of the whole world have fundamental economic consequences in the development of all countries. Our study combines financial elements which may influence the economic relations that currently exist between the countries of the world and the ones belonging to the European Union in particular. The processing of the high quantity of information is done through complex networks and the graph theory. They reflect the achievement and acceptance of new approaches related to the financial markets, but also to the most important macroeconomic indicator, GDP. Thus, we obtain new tools of analysis through methodological approaches on complex networks which allow numerous geometrical visualizations.

Optimizing business processes by learning from monitoring results

In the current paper we propose analyzing the recorded information about broken rules in business process execution and using it for enhancing business process models defined inside organizations for specific business cases. An alert system is designed as a Web Application for detecting deviations by verifying the event log data of the running cases. We consider it a Business Activity Monitoring solution implementing a custom rule definition language mainly created for ensuring the operational support inside organizations by providing guidance to people as task owners in real time about their actions not compliant with process definition. The current system is able to analyze real-time events with the scope of identifying problems, diagnosing them and generating alerts to recommend managerial action. The main effect should be a better performance of the organization and better business results. We consider that processes are meant to automatize and ease the work and should be commonly understood and accepted by the people contributing inside organizations. A rule constantly broken could indicate that a process difficult to follow in specific areas. Such statistical information about broken rules could represent the input for business process improvements. We also consider as a possibility referencing a process definition instead of a set of rules. The running cases are checked for conformance with the reference process model and if the case is not compliant, a trigger is activated and the information is used further for a possible evolution of business process definitions. The business process model is re-built from monitoring results and statistical information.

Collaborative Business Process Solution Considering an Ontological Dimension of Process Models

Considering the potential of collaborative business as a trend, we start from the assumption that collaborations between companies have a higher success rate if a good compatibility from process perspective is detected. The solution is based on the business processes used internally in all participating organizations to collaborations with common business objectives. For un-standardized organizations, we used process mining techniques to formalize processes and to attach them to business scenarios. The process models are reduced to graph format and graph comparison algorithms are used to compute a compatibility factor. Syntactic similarity is unfeasible for real business process comparison in an un-standardized environment. Semantic similarity is resource consuming and could lead to un-predictable results. We propose in the current paper a methodology based on domain specific ontology to calculate the similarity between organizations from process perspective and to offer a collaborative solution, business process based.