PERCUTANEOUS PCO2: A TECHNIQUE FOR CONTINUOUS MONITORING

By ANDREW WHITE

CONTENTS PART A EXPOSITION INTRODUCTION AND MOTIVATION PROBLEM STATEMENT AND RESEARCH OBJECTIVE THESIS STRUCTURE RESEARCH BACKGROUND BUSINESS PROCESS MANAGEMENT PROCESS DATA DATA-DRIVEN BUSINESS PROCESS MANAGEMENT APPROACHES DEEP LEARNING SUPERVISED MACHINE LEARNING FROM SHALLOW TO DEEP NEURAL NETWORKS ADVANCED DEEP NEURAL NETWORK ARCHITECTURES EXPLAINABLE ARTI?CIAL INTELLIGENCE FOR DEEP LEARNING DEEP LEARNING IN BUSINESS PROCESS MANAGEMENT DEEP LEARNING IN PROCESS ANALYSIS DEEP LEARNING IN PROCESS MONITORING RESEARCH APPROACH ADOPTED PARADIGM DESIGN SCIENCE RESEARCH PHILOSOPHICAL POSITIONING OF DESIGN SCIENCE RESEARCH RELEVANCE AND RIGOUR IN DESIGN SCIENCE RESEARCH IT ARTEFACTS AS CENTRAL ELEMENTS IN DESIGN SCIENCE RESEARCH ADOPTED DESIGN SCIENCE RESEARCH METHODOLOGY ADOPTED EVALUATION METHODS FOR DESIGN SCIENCE RESEARCH RESEARCH RESULTS A TECHNIQUE FOR PREDICTING NEXT CLICKS USING CONTEXT-AWARE LSTM NNS PROBLEM IDENTI?CATION AND DE?NITION OF OBJECTIVES ARTEFACT DESIGN DEMONSTRATION AND EVALUATION COMMUNICATION A TECHNIQUE FOR PREDICTING NEXT PROCESS ACTIVITIES USING GGNNS A TECHNIQUE FOR PREDICTING NEXT PROCESS ACTIVITIES AND TIMESTAMPS USING TIME-AWARE LSTM NNS A TECHNIQUE FOR EXPLAINABLE PROCESS OUTCOME PREDICTION USING GGNNS A TECHNIQUE FOR EXPLAINABLE NEXT ACTIVITY PREDICTION USING LSTM NNS AND LRP A TECHNIQUE FOR CREATING RELEVANCE-VALUE- AUGMENTED PROCESS MODELS USING GGNNS A TECHNIQUE FOR PREDICTING DEVIATIONS USING LSTM NNS A TECHNIQUE FOR DETECTING TEMPORAL WORKAROUNDS USING AES AND CNNS A TECHNIQUE FOR RECOMMENDING NEXT BEST ACTIONS USING LSTM NNS CLOSURE CONTRIBUTIONS LIMITATIONS FUTURE RESEARCH PART B THE STATE OF THE ART OF MACHINE LEARNING IN BUSINESS PROCESS MANAGEMENT: A SYSTEMATIC LITERATURE REVIEW INTRODUCTION BACKGROUND AND RELATED WORK RELATED BUSINESS PROCESS MANAGEMENT CONCEPTS RELATED MACHINE LEARNING CONCEPTS RELATED LITERATURE REVIEWS LITERATURE REVIEW APPROACH SEARCH PROCESS CODING PROCESS FUTURE RESEARCH ANALYSIS PROCESS RESULTS TYPE OF RESEARCH ML PROBLEMS, APPROACHES AND ALGORITHM GROUPS IN BPM LIFE-CYCLE PHASES AND USE CASES DISCUSSION IMPLICATIONS FUTURE RESEARCH DIRECTIONS CONCLUSION BRINGING LIGHT INTO THE DARKNESS - A SYSTEMATIC LITERATURE REVIEW ON EXPLAINABLE PREDICTIVE BUSINESS PROCESS MONITORING TECHNIQUES RELATED CONCEPTS RESEARCH METHOD CONCLUSION AND FURTHER STEPS A NEXT CLICK RECOMMENDER SYSTEM FOR WEB-BASED SERVICE ANALYTICS WITH CONTEXT-AWARE LSTMS RELATED WORK TOWARDS AN LSTM-BASED NEXT CLICK PREDICTOR PROBLEM STATEMENT OF?INE COMPONENT: THREE CON?GURATIONS TO LEARN AN LSTM ONLINE COMPONENT: APPLY AN LSTM FOR NEXT CLICK PREDICTION EVALUATION BASELINE WEB USAGE LOG EVALUATION PROCEDURE CONCLUSION AND FUTURE WORK EXPLORING GATED GRAPH SEQUENCE NEURAL NETWORKS FOR PREDICTING NEXT PROCESS ACTIVITIES BACKGROUND PRELIMINARIES GATED GRAPH SEQUENCE NEURAL NETWORKS FOR NEXT ACTIVITY PREDICTION THREE REPRESENTATION FORMS OF PROCESS INSTANCE GRAPHS LEARNING GATED GRAPH SEQUENCE NEURAL NETWORK MODELS EVENT LOGS PROCEDURE 10 TIME MATTERS: TIME-AWARE LSTMS FOR PREDICTIVE BUSINESS PROCESS MONITORING 10 1 INTRODUCTION 10 2 RELATED WORK 10 3 BACKGROUND 10 3 1 PRELIMINARIES 10 3 2 LONG SHORT-TERM MEMORY CELLS 10 4 METHODOLOGY 10 4 1 TIME-AWARE LONG SHORT-TERM MEMORY CELLS 10 4 2 NETWORK ARCHITECTURE 10 5 EXPERIMENTS 10 5 1 DATA SETS 10 5 2 PREPROCESSING 10 5 3 TRAINING SETUP 10 5 4 EVALUATION 10 5 5 IMPLEMENTATION 10 6 RESULTS 10 6 1 NEXT ACTIVITY PREDICTION 10 6 2 NEXT TIMESTAMP PREDICTION 10 7 DISCUSSION 10 8 CONCLUSION AND FUTURE WORK 11 EXPLAINABLE PREDICTIVE BUSINESS PROCESS MONITORING USING GATED GRAPH NEURAL NETWORKS 11 1 INTRODUCTION 11 2 BACKGROUND 11 3 RESEARCH METHOD 11 4 GATED GRAPH NEURAL NETWORKS FOR EXPLAINABLE OUTCOME PREDICTION IN BUSINESS PROCESSES 11 4 1 EVENT LOG TO GGNN INPUT TRANSFORMATION 11 4 2 GGNN-BASED PREDICTION MODEL CREATION 11 4 3 SINGLE GRAPH INSTANCE VISUALISATION 11 5 EVALUATION 11 5 1 DATA 11 5 2 TECHNICAL SETUP 11 5 3 RESULTS 11 5 4 DISCUSSION 11 6 CONCLUSION 12 XNAP: MAKING LSTM-BASED NEXT ACTIVITY PREDICTIONS EXPLAINABLE BY USING LRP 12 1 INTRODUCTION 12 2 BACKGROUND 12 2 1 PRELIMINARIES 12 2 2 LAYER-WISE RELEVANCE PROPAGATION FOR LSTMS 12 3 RELATED WORK ON EXPLAINABLE PBPM 12 4 XNAP: EXPLAINABLE NEXT ACTIVITY PREDICTION 12 4 1 OF?INE COMPONENT: LEARNING A BI-LSTM MODEL 12 4 2 ONLINE COMPONENT: PRODUCING PREDICTIONS WITH EXPLANATIONS 12 5 RESULTS 12 5 1 EVENT LOGS 12 5 2 EXPERIMENTAL SETUP 12 5 3 PREDICTIVE QUALITY 12 5 4 EXPLAINABILITY 12 6 CONCLUSION 13 A TECHNIQUE FOR DETERMINING RELEVANCE SCORES OF PROCESS ACTIVITIES USING GRAPH-BASED NEURAL NETWORKS 13 1 INTRODUCTION 13 2 BACKGROUND 13 2 1 EVENT LOGS 13 2 2 GRAPH NEURAL NETWORKS 13 2 3 GATED RECURRENT UNITS OF THE GATED GRAPH NEURAL NETWORK 13 3 GRM – DETERMINING RELEVANCE SCORES OF PROCESS ACTIVITIES WITH GGNNS 13 3 1 EVENT LOG TRANSFORMATION 13 3 2 GGNN MODEL CREATION AND TRAINING 13 3 3 PREDICTION AND RELEVANCE DETERMINING 13 4 EVALUATION 13 4 1 PROCEDURE 13 4 2 SETUP 13 4 3 DATA 13 4 4 RESULTS FOR PREDICTIVE QUALITY 13 4 5 CASE STUDY 13 5 DISCUSSION 13 6 CONCLUSION 14 PREDICTIVE BUSINESS PROCESS DEVIATION MONITORING 14 1 INTRODUCTION 14 2 BACKGROUND 14 2 1 CONFORMANCE CHECKING 14 2 2 PREDICTIVE BUSINESS PROCESS MONITORING AND DEEP LEARNING 14 3 RELATED WORK 14 4 RESEARCH METHOD 14 5 PREDICTIVE BUSINESS PROCESS DEVIATION MONITORING 14 5 1 PRELIMINARIES 14 5 2 DEVIATION LABEL 14 5 3 DATA REPRESENTATION TRANSFORMATION 14 5 4 MODEL LEARNING 14 5 5 MODEL APPLICATION 14 6 EVALUATION 14 6 1 EVENT LOGS 14 7 EVALUATION PROCEDURE 14 8 RESULTS 14 9 DISCUSSION 14 10CONCLUSION 15 DETECTING TEMPORAL WORKAROUNDS IN BUSINESS PROCESSES – A DEEP-LEARNING-BASED METHOD FOR ANALYSING EVENT LOG DATA 15 1 INTRODUCTION 15 2 THEORETICAL BACKGROUND 15 2 1 CONCEPTUAL FOUNDATIONS OF DETECTING WORKAROUNDS 15 2 2 PROCESS MINING FOR DETECTING WORKAROUNDS 15 3 RESEARCH METHOD 15 4 DEEP LEARNING FOR DETECTING WORKAROUNDS IN EVENT LOGS 15 4 1 PRELIMINARIES 15 4 2 SETTING UP AN INITIAL DATA SET 15 4 3 REMOVING NOISY DATA FROM THE DATA SET 15 4 4 ADDING WORKAROUNDS TO THE DATA SET 15 4 5 LEARNING A CLASSI?ER 15 4 6 APPLYING A CLASSI?ER 15 5 EVALUATION 15 5 1 EVENT LOGS 15 5 2 EVALUATION PROCEDURE 15 6 RESULTS 15 6 1 DETERMINING A THRESHOLD FOR NOISY DATA REMOVAL 15 6 2 TEST RESULTS FOR WORKAROUND DETECTION 15 6 3 RESULTS FOR WORKAROUND DETECTION 15 6 4 DEMONSTRATION OF WORKAROUND DETECTION 15 7 DISCUSSION 15 7 1 RESULTS 15 7 2 MANAGERIAL IMPLICATIONS 15 7 3 RELATED WORK 15 7 4 LIMITATIONS 15 8 CONCLUSION 16 PRESCRIPTIVE BUSINESS PROCESS MONITORING FOR RECOMMENDING NEXT BEST ACTIONS 16 1 INTRODUCTION 16 2 BACKGROUND 16 2 1 PRELIMINARIES 16 2 2 LONG SHORT-TERM MEMORY NEURAL NETWORKS 16 2 3 BUSINESS PROCESS SIMULATION 16 3 A PRBPM TECHNIQUE FOR RECOMMENDING NEXT BEST ACTIONS 16 3 1 OF?INE COMPONENT 16 3 2 ONLINE COMPONENT 16 4 EVALUATION 16 4 1 EVENT LOGS 16 4 2 PROCESS MODELS 16 4 3 PROCEDURE 16 4 4 RESULTS 16 5 DISCUSSION 16 6 RELATED WORK 16 7 CONCLUSION REFERENCES