Machine Learning for Predictive Modelling - …

1 1 2015 The MathWorks, Learning for Predictive ModellingRory Adams2 Machine Learning What is Machine Learning and why do we need it? Common challenges in Machine Learning Example: Human activity Learning using mobile phone data Example: Real-time object identification using images Example: Load forecasting using weather data Summary & Key TakeawaysAgenda3 Machine Learning is Everywhere Image Recognition Speech Recognition Stock Prediction Medical Diagnosis Data Analytics Robotics and [TBD]4 Machine LearningMachine Learning uses dataand produces a model to perform a taskStandard ApproachMachine Learning Approach = < >( _ , )ComputerProgramMachineLearning :Inputs OutputsHandWritten ProgramFormula or EquationIf X_acc> SITTING If Y_acc< 4 and Z_acc> 5then STANDING .. = 1 + 2 + 3 +..Task:Human Activity Detection :Predictors Response5 Different Types of LearningMachine LearningSupervised LearningClassificationRegressionUnsuperv ised Learning Discover a good internal representation Learn a low dimensional representation Response is a continuous number (temperature, stock prices).

2 Response is a choice between classes (True, False) (Red, Blue, Green)6 Example: Human Activity Learning Using Mobile Phone DataMachineLearningData: 3-axial Accelerometer data 3-axial Gyroscope data7 essentially, all models are wrong, but some are useful George Box8 StepsChallengeAccess, explore and analyzedataData diversityNumeric, Images,Signals, Text not always tabularPreprocess dataLack of domain toolsFiltering and feature extractionFeature selection and transformationTrain modelsTime consuming Train many models to find the best Assess model performanceAvoid pitfallsOver Fitting Speed-Accuracy-Complexity tradeoffsIterateChallenges in Machine LearningHard to get startedStepsChallengeAccess, explore and analyzedataData diversityNumeric, Images,Signals, Text not always tabularPreprocess dataLack of domain toolsFiltering and feature extractionFeature selection and transformationTrain modelsTime consuming Train many models to find the best Assess model performanceAvoid pitfallsOver Fitting Speed-Accuracy-Complexity tradeoffsIterate9 PREDICTIONM achine Learning WorkflowTrain: Iterate till you find the best modelPredict: Integrate trained models into applications MODELSUPERVISEDLEARNINGCLASSIFICATIONREG RESSIONPREPROCESS DATASUMMARYSTATISTICSPCAFILTERSCLUSTER ANALYSISLOAD DATAPREPROCESS DATASUMMARYSTATISTICSPCAFILTERSCLUSTER ANALYSISNEWDATAPREPROCESS DATASUMMARYSTATISTICSPCAFILTERSCLUSTER ANALYSISMODEL10 Machine Learning What is Machine Learning and why do we need it?

3 Common challenges in Machine Learning Example: Human activity Learning using mobile phone data Example: Real-time object identification using images Example: Load forecasting using weather data Summary & Key TakeawaysAgenda11 Example 1: Human Activity Learning Using Mobile Phone DataObjective: Train a classifier to classifyhuman activity from sensor dataData:Approach: Extract features from raw sensor signals Train and compare classifiers Test results on new sensor dataPredictors3-axial Accelerometer and Gyroscope dataResponseActivity:(Classification)12 PREDICTIONMODELM achine Learning Workflow for Example 1 Train: Iterate till you find the best modelPredict: Integrate trained models into applications MODELSUPERVISEDLEARNINGCLASSIFICATIONREG RESSIONPREPROCESS DATASUMMARYSTATISTICSPCAFILTERSCLUSTER ANALYSISLOAD DATAPREPROCESS DATASUMMARYSTATISTICSPCAFILTERSCLUSTER Machine Learning What is Machine Learning and why do we need it?

4 Common challenges in Machine Learning Example: Human activity Learning using mobile phone data Example: Real-time object identification using images Example: Load forecasting using weather data Summary & Key TakeawaysAgenda14 Example 2: Real-time Toy Identification Using ImagesObjective: Train a classifier to identify toytype from a webcam videoData:Approach: Extract features using Bag-of-words Train and compare classifiers Classify streaming video from a webcamPredictorsSeveral images of cars:ResponseCAR, HELICOPTER, PLANE, BIKE(Classification)15 PREDICTIONMODELM achine Learning Workflow for ExampleTrain: Iterate till you find the best modelPredict: Integrate trained models into applications MODELSUPERVISEDLEARNINGCLASSIFICATIONREG RESSIONPREPROCESS DATASUMMARYSTATISTICSPCAFILTERSCLUSTER ANALYSISLOAD DATAPREPROCESS DATASUMMARYSTATISTICSPCAFILTERSCLUSTER images as new featuresClassification LearnerEncode images as new features16 Machine Learning What is Machine Learning and why do we need it?

5 Common challenges in Machine Learning Example: Human activity Learning using mobile phone data Example: Real-time object identification using images Example: Load forecasting using weather data Summary & Key TakeawaysAgenda17 Example 3: Day-Ahead System Load ForecastingObjective: Train a neural network to predict the required system load for a zone Data:Approach: Extract additional features Train neural network Predict loadPredictorsTemperature, Dew point, Month, Day of week, Prior day load, Prior week loadResponseLOAD(Regression)18 PREDICTIONMODELM achine Learning Workflow for Example 1 Train: Iterate till you find the best modelPredict: Integrate trained models into applications MODELSUPERVISEDLEARNINGCLASSIFICATIONREG RESSIONPREPROCESS DATASUMMARYSTATISTICSPCAFILTERSCLUSTER ANALYSISLOAD DATAPREPROCESS DATASUMMARYSTATISTICSPCAFILTERSCLUSTER ANALYSISTESTDATATemp, Dew pointDay of weekPrior day loadPrior week loadNeural NetworkTemp, Dew pointDay of weekPrior day loadPrior week load19 Machine Learning What is Machine Learning and why do we need it?

6 Common challenges in Machine Learning Example: Human activity Learning using mobile phone data Example: Real-time object identification using images Example: Load forecasting using weather data Summary & Key TakeawaysAgenda20 StepsChallengeAccessing, exploring and analyzingdataData diversityPreprocess dataLack of domain toolsTrain modelsTime consuming Assess model performanceAvoid pitfallsOver Fitting,Speed-Accuracy-Complexity IterateChallenges in Machine Learning21 MATLAB Strengths for Machine LearningChallengeSolutionData diversityExtensive data supportImport and work with signal, images, financial, Textual, geospatial, and several others formatsLack of domain toolsHigh-quality librariesIndustry-standard algorithms for Finance, Statistics, Signal, Image processing & moreTime consuming Interactive,app-driven workflowsFocus on Machine Learning , not programingAvoid pitfallsOver Fitting,Speed-Accuracy-Complexity Integrated best practicesModel validation tools builtinto appRichdocumentation with step by step guidanceFlexible architecture for customized workflowsComplete Machine Learning platform22 Consider Machine Learning when: Hand written rules and equations are too complex Face recognition, speech recognition, recognizing patterns Rules of a task are constantly changing Fraud detection from transactions, anomaly in sensor data Nature of the data changes and the program needs to adapt Automated trading, energy demand forecasting, predicting shopping trends MATLAB for Machine LearningKey Takeaways23 Additional ResourcesDocumentation: Machine Learning with MATLAB:24Q & ATopic of interestSession / Demo StationWorkingwith IoTdataSession:MATLAB and the Internet of Things (IoT): Collecting and Analysing IoTDataAccessing, analysing and visualising dataSession.

7 Analysis of Experimental and Test DataWorkingwith big data setsSession:Tackling Big Data with MATLABD eploying machinelearning algorithmsDemo: Building MATLAB Apps to Visualise Complex DataMachinelearning with computer visionDemo: Identification of Objects in Real-Time Video