How To Evaluate and Improve Machine Learning Models
summary
Google Colab notebook walks through loading Titanic CSV, filling 177 missing ages with the mean, encoding gender, and building classifiers with NumPy/Pandas/Matplotlib and sklearn.svm; a custom 5‑fold cross‑validation computes per‑fold accuracies and shows linear SVM outperforms the kernel SVM, confirmed by a two‑sample z-score test at 95% confidence. Improvement steps include class balancing with RandomOverSampler (small accuracy lift) and an ensemble RandomForest (100 trees) that raises accuracy from ~78% to ~80%.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 3:19Why evaluating models matters
- 4:06Core performance metrics
- 5:51Cross validation for bias
- 7:37Methods to improve models
- 11:30Titanic dataset case study
- 17:40Loading Titanic dataset
- 25:15Implementing k-fold CV
- 32:09Evaluating SVM models
- 38:10Comparing models with z-score
- 44:24Improving models with oversampling
- 49:38Random forest classifier
- 51:16Summary
- 52:26Answering audience questions on model evaluation
related talks
27mUsing Machine Learning to solve a classification problem with scikit-learn
04h 03mTutorial - Intro to Data Science
51mHow to Test Artificial Intelligence and Machine Learning Solutions?
42mCreating correct and capable classifiers
01h 17mMachine Learning with Scikit-Learn
37mCreating correct and capable classifiers
01h 24mInpsect and try to interpret your scikit-learn machine-learning models
39mPredicting Titanic Survivors with Machine Learning