C.A.M.P.A.I.G.N Vijay Anam Partnership Performance Manager, Reading and Wokingham Local Police Areas © Vijay Anam, All Rights Reserved OUTLINE • • • • • • • INTRODUCTION NATIONAL INTELLIGENCE MODEL CAMPAIGN MODEL DATABASE MODULE GIS MODULE ADAPTIVE LAYER SUMMARY © Vijay Anam, All Rights Reserved INTRODUCTION • • • • CRIME ADAPTIVE MAPPING and PERFORMANCE POLICE REGISTERED SOCIAL LANDLORDS using • ARTIFICIAL • INTELLIGENCE and • GEOGRAPHICAL • NEIGHBOURHOODS © Vijay Anam, All Rights Reserved LOCAL AUTHORITY CRIME AND DISORDER REDUCTION PARTNERSHIP PROBATION SERVICE HEALTH AUTHORITY FIRE SERVICE NATIONAL INTELLIGENCE MODEL © Lancashire Constabulary 2000 © Vijay Anam, All Rights Reserved NATIONAL INTELLIGENCE MODEL – PROBLEM SOLVING MODEL • SCANNING – IDENTIFY THE PROBLEM(S) • ANALYSIS – DETERMINE CONTRIBUTORY FACTORS IN THE CONTEXT OF LOCATION LOCATION • LOCATION • VICTIM • OFFENDER • RESPONSE – PLAN AND INTITIATE INTERVENTIONS • ASSESSMENT – MONITOR PERFORMANCE OF THE INTERVENTION(S) – IDENTIFY SUCCESSES AND FAILURES – DEVELOP BEST VALUE © Vijay Anam, All Rights Reserved OFFENDER OFFENDER VICTIM VICTIM NATIONAL INTELLIGENCE MODEL – PROBLEM SOLVING MODEL - GIS ASSESS INTERVENTION(S) (ASSESSMENT - LAYER 3) DESIGN INTERVENTION(S) (RESPONSE - LAYER 3) IDENTIFY FACTORS (ANALYSIS - LAYER 2) IDENTIFY PROBLEMS (SCANNING - LAYER 1) Maps: © Crown Copyright, Account No.:100019672 © Vijay Anam, All Rights Reserved N.I.M – PROBLEM SOLVING MODEL – SCANNING • SCANNING – IDENTIFY HOTSPOTS – CREATE BOUNDARIES ENCIRCLING THE HOTSPOTS. – SELECT BOUNDRIES WHICH NEEDS URGENT ATTENTION • IN OUR EXAMPLE WE HAVE ENCIRCLED AN AREA WHICH NEEDS URGENT INTERVENTION – PROCEED TO THE NEXT STAGE © Vijay Anam, All Rights Reserved Maps: © Crown Copyright, Account No.:100019672 N.I.M – PROBLEM SOLVING MODEL – ANALYSIS • ANALYSIS (LIST ALL CONTRIBUTORY FACTORS) – IN CONTEXT OF LOCATION SUCH AS: • • • • • – SIMILARLY FROM THE POINT OF VIEW OF THE OFFENDER • • • • • – TYPES OF HOUSES AGE OF HOUSES DEPRIVATION GEOGRAPHY ….ETC. AGE OF OFFENDER SEX OF OFFENDER WHETHER L.P.O OR P.Y.O DRUG DEPENDENCY …..ETC. AND FINALY FROM THE VICTIM’S PERSPECTIVE • • • • AGE OF THE VICTIM SEX OF THE VICTIM WHETHER LIVING ALONE …..ETC. © Vijay Anam, All Rights Reserved Maps: © Crown Copyright, Account No.:100019672 AREA TO BE ANALYSED N.I.M – PROBLEM SOLVING MODEL – RESPONSE • RESPONSE – DESIGN INTERVENTIONS TO TACKLE PROBLEMS – SET TARGETS – SET MILESTONES – IDENTIFY LOCATIONS WHERE INTERVENTIONS WILL BE APPLIED © Vijay Anam, All Rights Reserved REGION WHERE INTERVENTIONS ARE PLANNED Maps: © Crown Copyright, Account No.:100019672 N.I.M – PROBLEM SOLVING MODEL – ASSESSMENT • BEFORE INTERVENTION ASSESSMENT – WHETHER MILESTONES WERE ON TRACK? – WHETHER TARGETS ACHIEVED ? – SUCCESSES – FAILURES – TYPES OF LESSONS LEARNT 10 9 8 7 6 5 4 3 2 1 0 AFTER INTERVENTION Target Actual Milestone1 MileStone2 © Vijay Anam, All Rights Reserved Milestone3 Final Traget CAMPAIGN MODEL ADAPTIVE GIS MODULE DATABASE MODULE MODULE © Vijay Anam, All Rights Reserved CAMPAIGN - NETWORK © Vijay Anam, All Rights Reserved CAMPAIGN MODEL – DATABASE MODULE SCANNING ANALYSIS RESPONSE ASSESSMENT PERFORMANCE PROJECT LOCATION INTERVENTIONS TARGETS PROBLEMS OFFENDER MILESTONES VICTIM TIME LINES © Vijay Anam, All Rights Reserved ASSESSMENT PERFORMANCE CAMPAIGN MODEL – GIS MODULE • • • • • • • POINT MAPS THEMATIC MAPS TIMELINE ANALYSIS TEMPORAL ANIMATION DEFINING RESPONSE BOUNDARIES AUTOMATIC PROJECT MONITORING QUERYING OF DATA LAYER © Vijay Anam, All Rights Reserved CAMPAIGN MODEL – ADAPTIVE MODULE PROBABILITY DISTRIBUTION REPORTING BAYESIAN NETWORK © Vijay Anam, All Rights Reserved ARTIFICIAL NEURAL NETWORKS STRUCTURE OF A BAYESIAN NETWORK A Bayesian Network is made up of two parts: 1) qualitative part • structure • Directed Acyclic Graph (DAG) • vertices represent random variables • edges represent relations between variables 2) quantitative part • the strength of relationship between variables • conditional probability function © Vijay Anam, All Rights Reserved P(N2|N1) P(N1) 1 2 3 4 P(N3|N1) P(N4|N3,N2) PURPOSE OF BAYESIAN NETWORKS • Models probabilities for any problem • Provides posterior beliefs in presence of some observations • Provides classification • Can explain why • Can find the variables with the most impact • Can model expert (subjective) knowledge • Learned from raw data © Vijay Anam, All Rights Reserved MODELLING A PROBLEM © Henrik Bengtsson © Vijay Anam, All Rights Reserved “The Year 2000 problem” ARTIFICIAL NEURAL NETWORK • AN ELECTRONIC MODEL OF THE FUNCTIONING OF THE HUMAN BRAIN – ON HOW TASKS OR FUNCTIONS ARE PERFORMED – BASED ON SIMPLE COMPUTATIONAL BUILDING BLOCKS • MAIN DESIGN CRITERIA: – CONNECTIONS BETWEEN NEURONS DETERMINE THE FUNCTION OF THE NETWORK – CAPACITY TO LEARN AND REMEMBER © Vijay Anam, All Rights Reserved TYPICAL NEURAL NETWORK Input Layer Weights Hidden Layer Weights x1 Input Signals x2 Output Target Output y1 t1 Error y2 xm Layer of Input Neurons © Vijay Anam, All Rights Reserved Layer of Hidden Neurons t2 Layer of Output Neurons © Matthew Casey SUMMARY • CAMPAIGN CAN EFFECTIVELY MONITOR PERFORMANCE OF ALL WORK UNDERTAKEN BY CRIME AND DISORDER REDUCTION PARTNERSHIPS • THE GIS MODULE IS USED AS AN INTERFACE TO REGISTER PROJECT BOUNDARIES AND PERFORM SPATIAL ANALYSIS. • DIFFERENT BAYESIAN NETWORKS (ADAPTIVE MODULE) TO ANALYSE SITUATIONS WHERE CAUSALITY PLAYS A ROLE AND UNDERSTANDING WHAT IS ACTUALLY GOING ON IS INCOMPLETE. • NEURAL NETWORKS (ADAPTIVE MODULE) USED FOR DETECTING PATTERNS AND FOR OPTIMISING DESIGN OF PROJECTS (PLANNING AND DELIVERY) © Vijay Anam, All Rights Reserved