elbowz / ml-hw3

Machine Learning: Missing Value forecast

Machine Learning: Missing Value forecast

more detailed information in ITALIAN: REPORT.MD

Compute probability and precision of hitting missing value through different models (MLP, SVM, RandomForest, KNN).

Utilized libs

• pandas: load and manipulation of DataSet
• numpy, collections: advanced data structures
• sklearn: algo, stats and splitting of TrainingSet
• argparse: command line script parsing
• matplotlib: graph plotting

Script usage

main.py [-h] [--test-size TEST_SIZE] [--question-marks-ts QM_REPETED_TS] [--no-split]
        [--img-tag IMG_TAG] [--verbose]
        [DATASET]

• DATASET: [breast-cancer|letters|poker] - select DataSet. Default is breast-cancer
• --test-size TEST_SIZE, -t TEST_SIZE: [0-1] - TestSet dimension. Default 0.2
• --question-marks-ts QM_REPETED_TS, -q QM_REPETED_TS: {0,1,2...} - Number of samples added to the DataSet for all '?' in the query. Default 1
• --no-split, -s: use the whole DataSet for training. Disabled by default
• --img-tag IMG_TAG, -i IMG_TAG: string - append a string to image graph filename
• --verbose, -v: verbose mode (increase it by repeating)
• -h, --help: show help

nota: Others options could be set by source code.

DataSets

The DataSet taken from archive.ics.uci.edu are:

• Breast Cancer Wisconsin
• Letter Recognition
• Poker Hand