IOT, timeseries and prediction with Android, Cassandra and Spark
summary
Android app posts smartphone accelerometer samples to a Spring Boot endpoint, Cassandra stores them as a time‑series using a Cassandra per-user-per-date schema (partition key user + date, timestamp clustering), Spark reads via the Spark‑Cassandra connector and Spark Streaming (custom Cassandra receiver), features are computed on 2-second windows and ~50 stats, and a Random Forest classifier yields ≈92–93% accuracy for walking/jogging/sitting/standing; demo runs end‑to‑end in real time.
AI-written summary · about
chapters (28)
- 0:00Intro
- 0:49Duchess France
- 2:32Transform data
- 3:13Accelerometer
- 4:08Collect data
- 4:58Cassandra Key Facts
- 5:09Timeseries with Cassandra
- 5:33Timeseries data model
- 6:03Storage model: Disk layout
- 7:06Query patterns
- 8:42Spark ecosystem
- 9:18Spark data sources
- 10:09Spark on cluster
- 10:49Spark Cassandra Connector
- 11:22Spark connection setup
- 11:51From Cassandra to Spark
- 12:38Spark Streaming
- 14:42Activity Recognition
- 15:31Multiclass classification
- 16:17Supervised learning
- 16:37Predictive model
- 17:29Android App
- 18:06Training data
- 18:38Prepare data windows
- 19:40Timeseries walking vs Jogging
- 20:13Features extraction
- 20:32Compute features
- 21:51Random forest
related talks
33mAnalytics in the age of the Internet of Things #bbuzz
42mSo, You Want to "Do Big Data?"
40mSmart cars of tomorrow: real-time driving patterns
19mBuild Real time, AI powered Apps
34mMining smartphone sensor data with python
01h 26mImplementing a highly scalable Stock prediction system with R, GemFire and Spring XD
30mSmart Streams: A Real-time framework for scoring Big and Fast Data
01h 48mBig Data made easy with a Spark (3.0)