퍼스널 빅데이터를 활용한 May 2016 마이닝 마인즈 핵심기술 개발 경희대학교 이 승 룡
Agenda 2 / • Evolution of Mining Minds • Mining Minds V2.5 Service Scenario • Mining Minds V2.5 Architecture • Platform Operations, Uniqueness and Contribution • Mining Minds V2.5 Microdemos • Future Plan • Conclusions
Evolution of Mining Minds 3
Mining Minds Platform Overview 4 / Behavior SERVICES Quantification Personalized Mining Minds Inputs Domain Expert Recommendation Real time Analytics Induced SERVICES Knowledge Habituation LLM Big data Privacy & Lifelog creation Storage Security End user Personalized well-being and health-care Platform Mining Minds Platform
History of Mining Minds 5 / MM V1.5 Personalized Lifestyle Coaching App MM V1.0 Personalized Lifestyle Coaching App MM V2.0 Weight Management App Rule Authoring Tool Rule Authoring Tool Management Tool May 2015 Behavior Inspection Tool Dec 2014 Behavior Inspection Tool Dec 2015 Real time activities simulation tool Data driven knowledge acquisition tool
History of Mining Minds 6 / Service API Service API Service Curation Layer Supporting Layer Service Orchastrator Input / Service Recommendation Manager Event Supporting Layer User Interface Security and Service Curation Layer Output Reasoning and Handler Curation Restful Adapter Privacy Recommendation Builder Recommendation Prediction Service User Interface Interpretor Data Service Curation Recommendation Manager UI Container Rule-based Service Curation Knowledge Context Explanation Access Reasoner Curation Webservice UI Container Service Contract Interface Interpreter Manager Reasoning Validator Service Recommendation Generator Data Client Curation Service Curation Service Explanation Generator Knowledge-base Data Knowledge Curation Layer Service Contract Data Client Curation Curation Service Knowledge Service Rule Builder Knowledgebase Client Acquisition Editor Client Information Curation Layer Oblivious Information Curation Layer Evaluator Low Level Context-Awareness Visualization Data Curation Low Level Context-Awareness Visualization Webservice Client Activity Recognizer fica Data Curation Parameter Client Descriptive Activity Recognizer Classi fica tion Analytics Data Curation Layer Data Curation Layer Visualization Sensory Data Processing Enabler Information Sensory Data Processing and Life-log Persistence Content and Persistence Modi fica tion Data Acquisition Representation Data Acquisition Data Curation and Reply fil Data Representation Query Restful Service Detection Data and Mapping Security and Service Data Low-level Creation Data Curation Service Data Curation Msg Curation Curation Context Interface Contract Preserver Privacy Send Recv. Data Curation Webservice Buf. Service Socket Server Preserver Service Contract CRUD Operations Data Model Client Rendering Life-log Representation and Mapping Trend DCL Object Model Analyzer Life-log Monitor Event Generator Credentials DCL Object CRUD Encrypted and Model Operations Personalized Big Data Intermediate Database Data Life-log Authorized Authorized Data Monitoring Big Data Socket Curation Big Data Storage Intermediate Database Object to Relational Curation Storage Server Object to Relational Service Processing User-Pro fil es Mapping Service Life-log Send / Recv. HDFS Mapping Client Client Life-log User-Pro fil es Messages Client HDFS HDFS Message Buffer UCLab Microsoft Azure Private Cloud Public Cloud Microsoft Azure Public Cloud MM V1.0 Primitive Data Curation (Representation and Mapping) MM V1.5 Big Data (only persistence) & Lifelog (representation and mapping) Smartphone-based Activity Recognition (5 activities) Smartphone + Smartwatch-based Activity Recognition (8 activities) Dec 2014 May 2015 Prebuilt Recommendations Generation (static) Rule-authoring tool (no executable knowledge) Static Admin Tool (raw data and activities) Situation-triggered Personalized Recommendations (local knowledge) Expert Inspection Tool (single/multi-user stats)
History of Mining Minds 7 / MM V2.0 Dec 2015 Big Data (only persistence) & Lifelog (representation and mapping) Smartphone + Smartwatch-based Activity Recognition (8 activities) Rule-authoring tool (no executable knowledge) Situation-triggered Personalized Recommendations (local knowledge) Expert Inspection Tool (single/multi-user stats) Real time activities monitoring (low level and corresponding high level) Data driven knowledge acquisition (utilizing big data approach)
Mining Minds Evolution 8 / Expert view Expert view Expert view End-user app view user app view user app view user app view PHYSICAL PHYSICAL PHYSICAL PHYSICAL + MM V 1.0 MM V 1.5 MM V 2.0 MM V 2.5 NUTRITION ACTIVITIES ACTIVITIES ACTIVITIES ACTIVITIES Knowledge base Physical SNS Trends UI/UX Activities RULE AUTHORING TOOL Physical Physical Physical Activities Activities Activities May 2015 Jan 2015 Dec 2015 May 2016
Mining Minds V2.5 Service Scenario 9
10 Overall Service Scenario for MM2.5 / Smart Smart Watch Phone Induced Domain Expert Habituation Physical Sensors Expert View Camera Expert Recommendation Kinect Personalized Logical Sensors Recommendation WARM SERVICES Personalized well-being and health-care Platform COLD SERVICES Twitter Inputs Facebook Human Behavior MM V3.0 Privacy & Security Quantification User Inputs lifelog Intermediate Database End user Big data Storage Expert
Mining Minds Platform and Services 11 / Service Requirements Expert-based Security and Privacy User Behavior Induced Habituation Personalized Services Aware Services Cold Services Services Quantification WARM SERVICES Sensors Warm Services Accelerometer • Educational • Anonymization • Physical • Expert direct • User aware Gyro • Oblivious • Situation facts recommendations activities • Gamification and education Matching aware analysis Audio Expert Services End User Services • New knowledge • Multimedia • Encryptions • Diet Control • Context Video (camera) evolution contents and Food Aware analysis GPS • Knowledge Authoring • Personalized • Lifeline Tool SNS recommendations • Knowledge creation and maintenance • Expert INPUTS Recommendations • User Behavior • User Behavior Expert Input Platform Requirement Inspection Tool Quantification • Visualization Context Multimodal Sensory Knowledge Acquisition Personalized UI/UX Knowledge (physical Determination BigData Processing & Maintenance Recommendations • Healthy habit activities & nutrition) induction • UI/UX Authoring Tool Direct • Data & Expert • Rule based • UI adaption rules • Real Time • Context • UI/UX • Security and Recommendations • Volume, Driven Reasoning fusioning Knowledge privacy • Situation • Context Knowledge velocity and Authoring User Input • Validation & • Descriptive variety ontology Evaluation • Life-log • Cross-domain • Semantic Verification Analytics Tagging • Knowledge • Security and representation recommendation Reasoing Engineering & Monitoring Privacy Taking Picture Toolkit Feedback
Cold Service Scenario-[1/2] 12 / OfficeWork, Amusement, Sleeping, Gardening, Inactivity, HouseWork, Exercising, Commuting, Snacks, Nuts, Vegetables, SeaFoods, Fruit, NoHLC Activity data Big data Storage Demographics High Level Context Preferences Risk Factors LLM Data Feedback Low Level Context synchronization Intermediate Database User Information Information Curation Data Curation Eating, LyingDown, Running, Sitting, Standing, Stretching, Sweeping, Walking Nutrition data Mining Minds Platform
Cold Service Scenario-[2/2] 13 / Information Curation High Level Context Low Level Context Service Curation Service Orchestrator Recommendation Builder 0 Big data Storage Recommendation Interpreter LLM Descriptive Data Analytics synchronization Intermediate Database Supporting Layer Data Curation Mining Minds Platform SNS
Warm Service Scenario-[1/4] 14 / Knowledge Authoring tool Rule Condition Condition Key Condition Key Is Situation! Condition Key Weight Status = Over Weight v Meal Time = Lunch v Food Intake = Cholesterol Enrich v Rule Condition Data Curation Condition Key Condition Key Condition Key Domain Model Meal Time = Dinner Manager Food Intake = Fiber Enrich Knowledge base Recommendation Lifelog Low cholesterol is recommended[Whole -wheat LLM Rule Builder Rule Validator Intermediate pasta with peas and spinach (Fiber enrich) ]and 20 minutes Brisk walk Database Knowledge Curation Big data Storage Save Rule Domain Expert View Expert Mining Minds Platform Knowledge Authoring Tool Big-data Descriptive Analytics
Warm Service Scenario-[2/4] 15 / Information Curation High Level Context Low Level Context Service Curation Service Orchestrator Recommendation Activity data Data Builder synchronization Monitoring Situation Demographics Recommendation Preferences Physical Interpreter Activity Risk Factors Monitoring Feedback Descriptive Analytics User Lifelog Lifelog Supporting Layer Information Monitoring Nutrition Monitoring Intermediate Database Data Curation Nutrition data Mining Minds Platform SNS
Recommend
More recommend
