Please enable JavaScript.

Coggle requires JavaScript to display documents.

Paper 3, Both algorithms can be used simultaneously. Its just the…

- - - - A priori
        
        Exploratory data processing
      - K-means clustering
  - - - Decision trees
        
        Predictive analytics
      - Support vector machines
        
        Image classification
      - Linear progressions
        
        Sales forecasting/Risk assessment
  - - - Active learning for user preference elicitation
        
        Q-learning
      - Personalized advertising
      - Learning optimal recommendation policies
- - - - Choosing "K"
        
        IF("K" = Too Small)
        
        Susceptible to outliers
        
        Have to go through to trial/error
        
        IF("K" = Too Large)
        
        a new point may be
        in a group that isn’t a specific enough match
        to make good recommendations (over smoothing)
    - - How to adjust the item/user feature matrix
        
        Cost Function
        
        Stochastic gradient decent
- - - - suggests using 80% of the data for training and 20% for validation
  - - - Precision - fraction of relevant instances among the retrieved instances
      - Recall - a measure of completeness - fraction of relevant instances that were reviewed
      - F-measure (F-score) - Balances precision and recall
      - Methods
        
        Mean Absolute Error (MAE) - measures the average magnitude of errors in a prediction - The average of the absolute difference between predicted and actual values - Better if there are many outliers (atypical data points)
        
        Root Mean Square (RMSE) - square root of the average of squared differences between prediction and actual observation