Machine Learning Outline - Githubissues

dssg / hitchhikers-guide

The Hitchhiker's Guide to Data Science for Social Good

997 stars 279 forks source link

Machine Learning Outline #87

Closed avishekrk closed 7 years ago

avishekrk commented 7 years ago

Outline

Explore the Data
- Previously univariate data

Clustering

Clusters over time windows see if we can group people

Define Features

Clusters over subset of features.

Outcomes

Recidivism 1 or 2 years

Build Classifiers

Models

In-sample testing
- why that is wrong

hold out set
go over cross-validation but going to do temporal cross validation

Evaluation

Accuracy
Precision-Recall
Confusion Matrix
Feature Importances

Feature Engineering

Pre-build features.