John Hopkins Practical Machine Learning
View the latset .docx (currently ml_course_project_v5_20160228.docx) for complete code and output of the predictive model.
You can also view the raw HTML output using the htmlpreview tool e.g.
http://htmlpreview.github.io/?https://github.com/nickkz/nk_jhu_ml/blob/master/ml_course_project_v4.html
The following steps were performed in developing the predictive model:
- initial workspace setup, load libraries, model tuning
- data loading and cleaning
- exploratory analysis to build intuition of important covariates
- reduce covariates through correlation analysis
- use 3 separate common machine learning algorithms (random forest, gradient boosting, linear discriminant analysis) to process data
- for each algo train model and predict test data.
- combine all 3 predictions into a common ensemble (stacked) model
- use this to determine final prediction model, apply to testing data set
Specific section comments and notes are inline below. Note: According to quiz results, this method resulted in 16 of 20 correct predictions, which was not a perfect result but good enough for passing.