1.2 - Intro to ML

• Computer learns the rules to answer questions
• A model is a function

1. Business problem
2. Data processing - explore, augment, feature engineering
3. Model - training, evaluation, interpretation
4. Meets business goal?
   1. Yes → Deploy → Collect more data for online training
   2. No → go back to data processing

ML Jargon

Label, Target, Output variable, "y"

Feature, Input variable, "x"

Feature engineering, transformation

Dimensionality

Model weight, parameters

Model training

The result to predict

Input data to make predictions

Reshaping raw input data to give more insights

Number of features

Applying optimization techniques to find best set of model weights

<aside> ⭐ ML focuses on prediction whereas stats focuses on interpretability

</aside>

1.3 - ML Applications

• Ranking - help users find the most relevant items
• Recommendation - give users items they may be interested in
• Classification - find what category an item belongs to
• Predicting a numerical value of an item - e.g. predicting sales for an ASIN
• Clustering - model figures out categories itself - categorize similar products together (e.g. finding duplicates across third-party sellers)
• Anomaly detection - finding uncommon items - e.g. Amazon Fresh finding decayed products

1.4 - Supervised and Unsupervised Learning