Science of Data

Science of Data

1. Readiness of data

   • Stage 0
   • Stage 1
   • Stage 2
   • Stage 3

2. Machine learning / Computational algorithms

   • Supervised learning
     • Classification
       • Training set data in pre-assigned groups
         • Assign new data to these groups
     • Regression
       • Use training set data to predict value of y for a new input
   • Unsupervised learning
     • K-means algorithm
       • Randomly initiates cluster centroids
         • Assigns data to each centroid
           • Centroids re-clustered until algorithm converges
       • Use elbow method for visual assessment
     • Hierachical clustering
       • Agglomerative
         • Bottom-up
           • Start with data as singletons
             • Merge clusters based on shortest distance
               • Algorithm stopped when data is merged into a single cluster
       • Divisive
         • Top-down
           • Data are single cluster
             • Algorithm partitions data into daughter clusters
               • Algorithm stops when data points are in singleton cluster
     • Principle component analysis
       • Used for dimensional reduction
         • Set directions that allow the greatest variability in the data
           • Repeat this until get the desireable number of dimensions
   • Reinforcement learning