Please enable JavaScript.

Coggle requires JavaScript to display documents.

Data Mining (process that uses statistical, mathematical and AI techniques…

- - - - If the prediction is a number (i.e. temperature 15°c) is a called a regression
    - - Metrics
        
        True negative rate (or sensitivity): TNR=TN/(TN+FP): Ratio correct negatives/ total negative count
        
        Accuracy: TP + TN/(TN+FP+TP+FN): Ratio correct instances/ total count of instances
        
        Precision: TP/(TP+FP) Ratio correct positives/ sum (in)correctly classified positives
        
        Recall: TP/(TP+FN) Ratio correct positives/ sum correctly classified positives + incorrectly classified negatives
        
        True positive rate (or sensitivity): TPR=TP/(TP+FN) Ratio correct positives/ total positive count
      - Simple split (holdout/ test sample estimation)
        
        Training set
        
        Model development #
        
        test set (or holdout set)
        
        Model assesment
        
        Prediction accuracy
      - k-fold cross validation
        
        Splits data set randomly into k mutually exclusive subset with ~equal size
        
        trains classification model k-1 times
        
        Accuracy = average k results against single 1 results
    - - Decision tree analysis
        
        1.Create root node
        
        Select best splitting attribute
        
        Splitting indices evaluate goodness
        
        Gini index
        
        Information gain
        
        Measures Entropy: extent of uncertainty or randomness
        
        Add a branch to the node for each value of the split
        
        Repeat step 2 and 3 until stopping criteria is reached/ the node is dominated by a single class label
      - Statistical analysis
        
        Logistic regression
        
        discriminant analysis
        
        Assumptions
        
        Normal distribution
  - - - Divisive: start wit one class, then brake appart
      - Agglomerative: start with individual classes, then join.
- - - - Nominal: codes assigned to objects as labels
      - Ordinal: codes assigned to objects as ranking labels
    - - Interval: eg temperature on celsius scale
      - Ratio: variables found in physics and engineering