Please enable JavaScript.

Coggle requires JavaScript to display documents.

DATA SCIENCE (basics (machine learning (how a machine 'learns'),…

- - - - simple
        
        one independent/explanatory variable
      - multiple
        
        more than one explanatory variable
      - relationships modeled using linear predictor models
        
        unknown parameters estimated from data
      - fitting
        
        least squares approach
        
        overall solution minimizes the sum of the squares
        of the residuals made in the results of each equation
        
        residuals: difference between an observed value and the fitted value from the model
        
        data fitting
        
        linear/ordinary
        
        occurs in statistical regression
        
        closed form solution
        
        nonlinear
        
        usually solved by iterative refinement
        
        system approximated by a linear one at each iteration
    - - DV is categorical
        
        binary
        
        estimate probability of a binary response
        based on 1+ predictors
        
        multinominal
        
        outcome can have 3+ possible types
        
        ordinal
        
        DV's that are ordered
      - predicting DV's that take membership in 1 of a limited number of categories
      - analogous to linear regression, but violates some of its basic assumptions
        
        residuals cannot be normally distributed
      - process
        
        need to convert binary variable into a continuous one
        
        take the odds of the event happening for different levels of each IV
        
        rake the ratio of those odds
        
        take the logarithm of that ratio - called the logit
        
        logit fitted to the predictors using linear regression analysis
      - measures relationship b/w categorical DV and 1+ IV's
        
        by estimating probabilities using a logistic function
    - - between a dependent variable and 1 IV('predictors')
      - understand how typical value of DV changes when any of the IV is varied
    - - prediction and forecasting
      - determining which if the IV is related to the DV
  - - - non-parametric method
      - used for classification and regression
      - input: the k closest training examples in the feature space
      - output: depends on whether doing classification or regression
        
        classification: class membership
        
        object classified by majority vote of its neighbors
        
        regression
        
        property value for the object which is the average of its k nearest
      - instance-based/lazy learning
        
        function is only approximated locally, all computation deferred until classification