Predictive Model Exploration Tool of Car Evaluation Data Set

• Car Evaluation Data Set
  • https://archive.ics.uci.edu/ml/datasets/Car+Evaluation
    
• We have to explore multiple predictive models
  • Generalized Boosted Models (gbm)
  • Linear Discriminant Analysis (lda)
  • Random Forest (rf)
  • Recursive Partitioning (rpart)
• We also have to consider various sizes of the training data set
• Some predictive models take time to compute

• Manual compilation of model accuracy of various models with various sizes of traing data set
  • very time consuming
  • error-prone

  percent   gbm   lda    rf rpart
1      60 97.83 87.10 95.80 83.77
2      61 97.62 88.97 96.87 85.39
3      62 97.40 89.60 96.48 84.10
4      63 98.12 88.87 96.24 81.97
5      64 96.61 89.19 95.65 87.90
6      65 96.52 88.23 96.35 83.75

• A Shiny App: Predictive Model Exploration Tool
• What it offers:
  • Confusion matrix and statistics of a selected model
    • To improve app performance
      • Pre-calculate gbm model fits
      • Store model fits as object files on server
  • Training data set of selected size
  • Model accuracy in the form of a chart
    • Easy to compare the accuracy
  • Last but not lease: Efficiency and Accuracy!!