Skip to content

Latest commit

 

History

History
71 lines (54 loc) · 2.19 KB

history.md

File metadata and controls

71 lines (54 loc) · 2.19 KB

03/12

Assigniment of methods:

  • Gaurav: Decision Tree Classifier
  • Mert: Random Forests
  • Giacomo: Boosting
  • Stefania: Stacking
  • Martina: Bagging

07/12

Call summary

  • We have all explained the assigned method.
  • We have set a new date for the meeting to conclude this "documentation" part on 10/12 after sds test
  • For the next call we would like to decide how to divide tasks for the paramaters optimization, code implementation ...

Methods

Summary on methods in methods_documentation/

New notebook

  • Updated deprecated libraries and functions
  • Deleted unused tuning fucntions
  • Modified notebook is file machine-learning-for-mental-health-1.ipynb

Proposed modification

  • Tuning parameters
  • Include country info with new dataset
  • Change barplot to piechart

10/12

Tasks:

Tuning parameters

  • Mert and Gaurav: Random Forest and Tree classifiers
  • Stefania and Martina linear regression and K-neighbors
  • Giacomo ensemble methods

Cross validation

  • Mert, Gaurav,Martina and Stefania: Apply cross correlation on ML methods

Select features

  • Martina and Stefania: Pick the most correlated parameters for the analysis
Decision trees Random forest Stacking Bagging Boosting Knn Regression
Tuning Tuning Tuning Tuning Tuning Tuning Tuning
Cross correlation Cross correlation Test metaclassifier - - Cross correlation Cross correlation
- - - - - - Adapt evaluation

Summary on what we have done so far (feel free to add)

  • Updated deprecated libraries and functions
  • Deleted unused functions and neural network analysis
  • Added country happiness index column
  • Changed the list of columns used as parameters for the classification: from random cols to higly correlated cols
  • Adapted our code for regression to this dataset
  • Documentation on all methods
  • Improoved some visualizations
  • Tested random forest with different parameters

18/12

New meeting to discuss final resluts on tuning and decide how to visualize this information

20-21/12

Final meeting for the slides