Amazon Machine Learning to Predict Responses
summary
A prototype called Suspicious Activity Monitor (S.A.M.) uses Amazon Machine Learning, Amazon Rekognition, AWS Lambda and S3 to classify photos+metadata (age, gender, county, time, day) into a binary crime/no-crime prediction trained on Georgia Bureau of Investigation records. Model training uses a 70/30 train/test split, returns AUC 0.897, and exposes a real‑time endpoint (small fees: ~$1.12 to build, ~$11/month idle in examples). Demo extracts vision attributes, calls the model, and highlights the false positive/false negative tradeoff and risks of machine bias.
AI-written summary · about
chapters (44)
- 0:00Intro
- 4:36About Me
- 8:21General Uses of Machine Learning
- 10:38When to Use Machine Learning
- 11:23What Machine Learning is Not
- 12:36When NOT to Use Machine Learning
- 13:06Predictive Policing
- 14:06Study by University of Pennsylvania
- 14:30United Kingdom (UK) Metropolitan Police
- 15:12Pittsburgh, Pennsylvania Police Department
- 15:41Chicago Police Department (CPD)
- 22:22How Machines Learn
- 23:52Binary
- 24:32Multiclass
- 25:05Regression
- 25:35Unsupervised Learning
- 26:23Answerable Questions
- 27:10Reinforcement Learning
- 27:59The Machine Learning Process
- 30:10Obtain Data
- 32:22Crime Conclusions Based on Analysis
- 33:12Prepare & Load Data
- 35:09Prepare Data
- 36:05Sample Data Observation
- 36:55Lessons Learned
- 38:07Load the Data
- 40:59Create Datasource
- 41:22Datasource Mapping Exercise
- 41:54Datasource Created
- 42:34Sanity Check Data (cont.)
- 43:16Create (Train) Model (cont.)
- 44:13Learning Algorithm
- 45:33Linear Models
- 51:03Enable the Model for Predictions
- 51:28Enable Model for Real-Time Predictions
- 53:20SAM Front End
- 55:20Main Event Handler
- 56:10Preparing the Response
- 56:36Retrieving Crime Prediction
- 57:16Test Results from Lambda Console
- 57:45Deleting a Real-Time Endpoint
- 57:57Things to Note
- 58:26Machine Blas
- 59:40Future of Machine Learning
related talks
24mAccelerate your business with AWS AI Services
55mEthical Machine Learning
39mFraud Detection: Using ML to Identify and Manage Fraudulent Activities
01h 41mGetting started with Machine Learning on AWS.
55mMachine Learning and AI in Java on AWS
42mHow to Build Smart Applications Without a Machine Learning Background
51mProfundizando en Aprendizaje Automático para Desarrolladores
31mCatch more potential online fraud faster with Amazon Fraud Detector