PhD Dissertation

Data Preprocessing

All relevant Python scripts are in the directory Scripts/Preprocessing/Python.

• Exploratory data analysis (EDA): describe.py
• Data encoding: convert.py
• Metadata summary: metasum.py
• Setting number of cuts: setcut.py
• Sampling features including instances through SelectKBest: selectkbest.py
• Sample EDA: sampledesc.py

Decision Tree Classifier

• Scripts/ML/Python/dtree.py

Proposed Mixed Box Classifier

• The most accurate box classifer is built by the CPLEX optimizer.
• The OPL project locates at Projects.
• Important parameters
  • cplex.intsollim: MIP solution number limit (default: 9223372036800000000)
  • cplex.threads: parallel threads (default: 0 implying up to 32 threads)
  • cplex.workmem: working memory before compression and swap (in MB) (default: 2048)
  • cplex.trelim: uncompressed tree memory limit (in MB) (default: 1e+75)
  • cplex.nodefileind: node storage file switch
    • 0 = No node file
    • 1 = Node file in memory and compressed (default)
    • 2 = Node file on disk
    • 3 = Node file on disk and compressed (used in this work)
  • cplex.status: solution status
    • Used for exiting
      • 1 = CPX_STAT_OPTIMAL
      • 101 = CPXMIP_OPTIMAL
      • 102 = CPXMIP_OPTIMAL_TOL
      • 111 = CPXMIP_MEM_LIM_FEAS (Predefined tree memory limit is exceeded and a solution is feasible)
      • 112 = CPXMIP_MEM_LIM_INFEAS
    • Not used for exiting in this work
      • 11 = CPX_STAT_ABORT_TIME_LIM (Accumulated computational time is already recorded in each iteration)
      • 104 = CPXMIP_SOL_LIM (All solutions with different objective values are recorded)
• Execution: oplrun -p box 2>&1 | tee <LOG_FILEPATH>
• To reorder the decision regions based on the cuts obtained by the OPL model and to find the complete sets of their predicted class labels, execute Python script file Scripts/Box/Python/finalbox.py