###
###  UCI MAGIC Gamma Telescope Data
###

Local copy:
  magic.csv

Downloaded From:
  https://www.openml.org/d/1120
  (CSV file: https://api.openml.org/data/v1/get_csv/54003)

Original Source:
  UCI Machine Learning Repository
  https://archive.ics.uci.edu/dataset/159/magic+gamma+telescope

Variable Information:

  Name -- Data Type -- Measurement -- Description
  ID        -- not a feature. should be removed.
  fLength   -- quantitative -- mm            -- major axis of ellipse
  fWidth    -- quantitative -- mm            -- minor axis of ellipse
  fSize     -- quantitative -- log10(#phot)  -- 10-log of sum of content of all pixels
  fConc     -- quantitative -- ratio         -- ratio of sum of two highest pixels over fSize
  fConc1    -- quantitative -- ratio         -- ratio of highest pixel over fSize
  fAsym     -- quantitative -- mm            -- distance from highest pixel to center, projected onto major axis
  fM3Long   -- quantitative -- mm            -- 3rd root of third moment along major axis
  fM3Trans  -- quantitative -- mm            -- 3rd root of third moment along minor axis
  fAlpha    -- quantitative -- deg           -- angle of major axis with vector to origin
  fDist     -- quantitative -- mm            -- distance from origin to center of ellipse
  class     -- binary (g, h) -- none         -- g = gamma (signal), h = hadron (background) -- response variable

Summary:
  Number of instances (observations): 19020
  Number of Attributes: 11
  Attribute breakdown: 10 quantitative feature variables, and 1 binary response variable
  Missing Attribute Values: None
  Class distribution: g (gamma, signal) = 12332 (64.8%), h (hadron, background) = 6688 (35.2%)


Abstract:
  Data are Monte Carlo generated to simulate the registration of high energy
  gamma particles in an atmospheric Cherenkov telescope. The task is to
  discriminate images caused by primary gamma rays (signal) from images of
  hadronic showers initiated by cosmic rays in the upper atmosphere (background).

Data Characteristics:
  The data were generated by the Monte Carlo program CORSIKA, run with parameters
  allowing events with energies down to below 50 GeV. Each shower image is
  pre-processed into an elongated cluster, and a principal component analysis in
  the camera plane defines an ellipse. The attributes are the characteristic
  (Hillas) parameters of this ellipse plus further image descriptors. The values
  are in raw form (not scaled).

Notes:
  The OpenML CSV file also has a first column named ID, a row identifier that is
  not a feature, and its column names end in a colon (for example "fLength:").
  The R code below drops ID and removes the colons.
  For technical reasons the number of h events is underestimated; in the real
  data the h class is the majority. Plain classification accuracy is therefore
  not meaningful, because accepting a background event as signal is worse than
  rejecting a signal event. The data providers recommend comparing classifiers
  with an ROC curve, at the points where the probability of accepting a
  background event as signal is below 0.01, 0.02, 0.05, 0.1 or 0.2.

Reference:
  Bock, R.K., Chilingarian, A., Gaug, M., Hakl, F., Hengstebeck, T., Jirina, M.,
  Klaschka, J., Kotrc, E., Savicky, P., Towers, S., Vaicilius, A., Wittek W. (2004).
  Methods for multidimensional event classification: a case study using images
  from a Cherenkov gamma-ray telescope. Nucl. Instr. Meth. A, 516, pp. 511-528.


----------
library(tidyverse)
DF <- read_csv("magic.csv", check.names=FALSE)     # keep names exactly as in the file
DF1 = DF %>%
    select(-ID) %>%                                # drop the column "ID"
    rename_with(~ gsub(":", "", .x))               # remove ":" from column names

DF1