Life cycle of data becoming a model
summary
Starts from three concrete data types — behavioral, feedback, synthetic — and names the tooling and tradeoffs: build secure pipelines with Spark and Hadoop, ensure randomized sampling to avoid collection bias, or account for it later. Feature engineering follows profiling, aggregation, density and contextual features with binning and log transforms. Models favor generative methods with small data, use random forest bootstrapping for robustness, handle class imbalance via down/up‑sampling or class weights, and evaluate with precision/recall, AUC and macro metrics plus live monitoring.
AI-written summary · about
chapters (21)
- 0:00Intro
- 0:31Motive
- 1:31Cycle of Development
- 2:18Sources of Data Collection & Examples
- 2:56Behaviour data - common examples
- 3:30Behaviour data collection - Skills Involved
- 4:38Data Ingestion -science problems
- 5:48Big Data - science problems
- 6:43Feedback data - common examples
- 7:17Feedback data - Outline
- 9:58Synthetic data - Common Examples
- 10:20Synthetic data - Outline
- 13:33Example - Behaviour data features
- 14:39Examples: Density Features
- 15:31Histogram study
- 19:14Log transformation (before & after)
- 20:22Common questions - ML model training
- 21:50Bootstrapping
- 22:56Imbalanced classes - Classification problem
- 24:02Penalty computation is modified
- 28:57Macro Accuracy - Preferences
related talks
48mThe Missing Step in Moving from Data to Models
39mDoing Data: The Critical Process of Data Preparation (...)
26mTesting a Data Science Model
37mLife of a data scientist
26mUnlocking the Power of Data-Centric AI.
37mBig Data Analytics, Machine Learning, Real Time Processing
17mProductionizing Data Science Workloads
25m"Data Engineering: The World Between Worlds"