Applied Machine Learning

Model = Algorithm ( Parameters = Values ) + Data

Essential Steps

1. Explore the Data

• Data Type
• Information
• Rows and Columns
• Numerical or Categorical
• Find Missing Data
• Find Relation between Features and Labels
• Visual Data

2. Clean the Data

• Fill or Drop Missing
• Encode Categorical Data

3. Split Datset into Train Set, Validate Set and Test Set

4. Fit | Train an Initial Model and Evaluate

• Use K Fold Cross Validation to get Better Accuracy and Observe the Cross Validation Score

5. Tune Hyperparameters by using Grid Search Cross Validation

• Apply Grid Search Cross Validation to Find Optimal Hyperparameters of a Model which results in the most Accurate Predictions
• Find Best Parameters

6. Evaluate on Validation Set

• Evaluate the Results on Validation Set using the Best Performing Parameters
• Create more than one Model to Find Best Performing Model for Test Set

7. Select and Evaluate the Final Model on Test Set

• Select the Final Best Performing Model on Test Set for Evaluation.

Model	Type	Train Speed	Predict Speed	Performance
Logistic Regression	Classification	Fast	Fast	Low
Support Vector Machine	Classification	Slow	Moderate	Medium
Multi Layer Perceptron	Both	Slow	Moderate	High
Random Forest	Both	Moderate	Moderate	Medium
Boosted Tree	Both	Slow	Fast	High