Click through the tabs to explore our results

Setup

library(tidyverse)
library(caret)
library(skimr)

Data

training <- read_csv("https://d396qusza40orc.cloudfront.net/predmachlearn/pml-training.csv",
                    na = c("NA", "#DIV/0!", ""))
testing <- read_csv("https://d396qusza40orc.cloudfront.net/predmachlearn/pml-testing.csv",
                    na = c("NA", "#DIV/0!", ""))

training %>% 
  skimr::skim()
Data summary
Name Piped data
Number of rows 19622
Number of columns 160
_______________________
Column type frequency:
character 4
logical 6
numeric 150
________________________
Group variables None

Variable type: character

skim_variable n_missing complete_rate min max empty n_unique whitespace
user_name 0 1 5 8 0 6 0
cvtd_timestamp 0 1 16 16 0 20 0
new_window 0 1 2 3 0 2 0
classe 0 1 1 1 0 5 0

Variable type: logical

skim_variable n_missing complete_rate mean count
kurtosis_yaw_belt 19622 0 NaN :
skewness_yaw_belt 19622 0 NaN :
kurtosis_yaw_dumbbell 19622 0 NaN :
skewness_yaw_dumbbell 19622 0 NaN :
kurtosis_yaw_forearm 19622 0 NaN :
skewness_yaw_forearm 19622 0 NaN :

Variable type: numeric

skim_variable n_missing complete_rate mean sd p0 p25 p50 p75 p100 hist
…1 0 1.00 9811.50 5664.53 1.00000e+00 4906.25 9811.50 14716.75 1.962200e+04 ▇▇▇▇▇
raw_timestamp_part_1 0 1.00 1322827119.27 204927.68 1.32249e+09 1322673099.00 1322832920.00 1323084264.00 1.323095e+09 ▃▃▇▁▆
raw_timestamp_part_2 0 1.00 500656.14 288222.88 2.94000e+02 252912.25 496380.00 751890.75 9.988010e+05 ▇▇▇▇▇
num_window 0 1.00 430.64 247.91 1.00000e+00 222.00 424.00 644.00 8.640000e+02 ▇▇▇▇▇
roll_belt 0 1.00 64.41 62.75 -2.89000e+01 1.10 113.00 123.00 1.620000e+02 ▇▁▁▅▅
pitch_belt 0 1.00 0.31 22.35 -5.58000e+01 1.76 5.28 14.90 6.030000e+01 ▃▁▇▅▁
yaw_belt 0 1.00 -11.21 95.19 -1.80000e+02 -88.30 -13.00 12.90 1.790000e+02 ▁▇▅▁▃
total_accel_belt 0 1.00 11.31 7.74 0.00000e+00 3.00 17.00 18.00 2.900000e+01 ▇▁▂▆▁
kurtosis_roll_belt 19226 0.02 -0.22 2.92 -2.12000e+00 -1.33 -0.90 -0.22 3.300000e+01 ▇▁▁▁▁
kurtosis_picth_belt 19248 0.02 4.33 11.58 -2.19000e+00 -1.11 -0.15 3.18 5.800000e+01 ▇▁▁▁▁
skewness_roll_belt 19225 0.02 -0.03 0.92 -5.74000e+00 -0.44 0.00 0.42 3.600000e+00 ▁▁▆▇▁
skewness_roll_belt.1 19248 0.02 -0.30 2.26 -7.62000e+00 -1.11 -0.07 0.66 7.350000e+00 ▁▂▇▁▁
max_roll_belt 19216 0.02 -6.67 94.59 -9.43000e+01 -88.00 -5.10 18.50 1.800000e+02 ▇▅▁▁▃
max_picth_belt 19216 0.02 12.92 8.01 3.00000e+00 5.00 18.00 19.00 3.000000e+01 ▇▁▆▃▁
max_yaw_belt 19226 0.02 -0.22 2.92 -2.10000e+00 -1.30 -0.90 -0.20 3.300000e+01 ▇▁▁▁▁
min_roll_belt 19216 0.02 -10.44 93.62 -1.80000e+02 -88.40 -7.85 9.05 1.730000e+02 ▁▇▅▁▃
min_pitch_belt 19216 0.02 10.76 7.47 0.00000e+00 3.00 16.00 17.00 2.300000e+01 ▇▁▁▇▂
min_yaw_belt 19226 0.02 -0.22 2.92 -2.10000e+00 -1.30 -0.90 -0.20 3.300000e+01 ▇▁▁▁▁
amplitude_roll_belt 19216 0.02 3.77 25.26 0.00000e+00 0.30 1.00 2.08 3.600000e+02 ▇▁▁▁▁
amplitude_pitch_belt 19216 0.02 2.17 2.36 0.00000e+00 1.00 1.00 2.00 1.200000e+01 ▇▁▁▁▁
amplitude_yaw_belt 19226 0.02 0.00 0.00 0.00000e+00 0.00 0.00 0.00 0.000000e+00 ▁▁▇▁▁
var_total_accel_belt 19216 0.02 0.93 2.22 0.00000e+00 0.10 0.20 0.30 1.650000e+01 ▇▁▁▁▁
avg_roll_belt 19216 0.02 68.06 63.14 -2.74000e+01 1.10 116.35 123.38 1.574000e+02 ▇▁▁▃▆
stddev_roll_belt 19216 0.02 1.34 2.44 0.00000e+00 0.20 0.40 0.70 1.420000e+01 ▇▁▁▁▁
var_roll_belt 19216 0.02 7.70 23.16 0.00000e+00 0.00 0.10 0.50 2.007000e+02 ▇▁▁▁▁
avg_pitch_belt 19216 0.02 0.52 22.41 -5.14000e+01 2.02 5.20 15.78 5.970000e+01 ▃▁▇▃▁
stddev_pitch_belt 19216 0.02 0.60 0.64 0.00000e+00 0.20 0.40 0.70 4.000000e+00 ▇▁▁▁▁
var_pitch_belt 19216 0.02 0.77 1.76 0.00000e+00 0.00 0.10 0.50 1.620000e+01 ▇▁▁▁▁
avg_yaw_belt 19216 0.02 -8.83 93.48 -1.38300e+02 -88.18 -6.55 14.12 1.735000e+02 ▇▁▅▁▃
stddev_yaw_belt 19216 0.02 1.34 10.29 0.00000e+00 0.10 0.30 0.70 1.766000e+02 ▇▁▁▁▁
var_yaw_belt 19216 0.02 107.49 1655.52 0.00000e+00 0.01 0.09 0.48 3.118324e+04 ▇▁▁▁▁
gyros_belt_x 0 1.00 -0.01 0.21 -1.04000e+00 -0.03 0.03 0.11 2.220000e+00 ▁▇▁▁▁
gyros_belt_y 0 1.00 0.04 0.08 -6.40000e-01 0.00 0.02 0.11 6.400000e-01 ▁▁▇▁▁
gyros_belt_z 0 1.00 -0.13 0.24 -1.46000e+00 -0.20 -0.10 -0.02 1.620000e+00 ▁▂▇▁▁
accel_belt_x 0 1.00 -5.59 29.64 -1.20000e+02 -21.00 -15.00 -5.00 8.500000e+01 ▁▁▇▁▂
accel_belt_y 0 1.00 30.15 28.58 -6.90000e+01 3.00 35.00 61.00 1.640000e+02 ▁▇▇▁▁
accel_belt_z 0 1.00 -72.59 100.45 -2.75000e+02 -162.00 -152.00 27.00 1.050000e+02 ▁▇▁▅▃
magnet_belt_x 0 1.00 55.60 64.18 -5.20000e+01 9.00 35.00 59.00 4.850000e+02 ▇▁▂▁▁
magnet_belt_y 0 1.00 593.68 35.68 3.54000e+02 581.00 601.00 610.00 6.730000e+02 ▁▁▁▇▃
magnet_belt_z 0 1.00 -345.48 65.21 -6.23000e+02 -375.00 -320.00 -306.00 2.930000e+02 ▁▇▁▁▁
roll_arm 0 1.00 17.83 72.74 -1.80000e+02 -31.78 0.00 77.30 1.800000e+02 ▁▃▇▆▂
pitch_arm 0 1.00 -4.61 30.68 -8.88000e+01 -25.90 0.00 11.20 8.850000e+01 ▁▅▇▂▁
yaw_arm 0 1.00 -0.62 71.36 -1.80000e+02 -43.10 0.00 45.88 1.800000e+02 ▁▃▇▃▂
total_accel_arm 0 1.00 25.51 10.52 1.00000e+00 17.00 27.00 33.00 6.600000e+01 ▃▆▇▁▁
var_accel_arm 19216 0.02 53.24 53.98 0.00000e+00 9.03 40.61 75.62 3.317000e+02 ▇▂▁▁▁
avg_roll_arm 19216 0.02 12.68 68.58 -1.66670e+02 -38.37 0.00 76.32 1.633300e+02 ▁▅▇▅▃
stddev_roll_arm 19216 0.02 11.20 17.10 0.00000e+00 1.38 5.70 14.92 1.619600e+02 ▇▁▁▁▁
var_roll_arm 19216 0.02 417.26 2007.16 0.00000e+00 1.90 32.52 222.65 2.623221e+04 ▇▁▁▁▁
avg_pitch_arm 19216 0.02 -4.90 26.83 -8.17700e+01 -22.77 0.00 8.28 7.566000e+01 ▁▃▇▂▁
stddev_pitch_arm 19216 0.02 10.38 9.40 0.00000e+00 1.64 8.13 16.33 4.341000e+01 ▇▅▂▁▁
var_pitch_arm 19216 0.02 195.86 292.60 0.00000e+00 2.70 66.15 266.58 1.884560e+03 ▇▁▁▁▁
avg_yaw_arm 19216 0.02 2.36 61.33 -1.73440e+02 -29.20 0.00 38.18 1.520000e+02 ▁▂▇▃▂
stddev_yaw_arm 19216 0.02 22.27 23.69 0.00000e+00 2.58 16.68 35.98 1.770400e+02 ▇▂▁▁▁
var_yaw_arm 19216 0.02 1055.93 2722.17 0.00000e+00 6.64 278.31 1294.85 3.134457e+04 ▇▁▁▁▁
gyros_arm_x 0 1.00 0.04 1.99 -6.37000e+00 -1.33 0.08 1.57 4.870000e+00 ▁▃▇▆▂
gyros_arm_y 0 1.00 -0.26 0.85 -3.44000e+00 -0.80 -0.24 0.14 2.840000e+00 ▁▂▇▂▁
gyros_arm_z 0 1.00 0.27 0.55 -2.33000e+00 -0.07 0.23 0.72 3.020000e+00 ▁▂▇▂▁
accel_arm_x 0 1.00 -60.24 182.04 -4.04000e+02 -242.00 -44.00 84.00 4.370000e+02 ▇▅▇▅▁
accel_arm_y 0 1.00 32.60 109.87 -3.18000e+02 -54.00 14.00 139.00 3.080000e+02 ▁▃▇▆▂
accel_arm_z 0 1.00 -71.25 134.65 -6.36000e+02 -143.00 -47.00 23.00 2.920000e+02 ▁▁▅▇▁
magnet_arm_x 0 1.00 191.72 443.64 -5.84000e+02 -300.00 289.00 637.00 7.820000e+02 ▆▃▂▃▇
magnet_arm_y 0 1.00 156.61 201.91 -3.92000e+02 -9.00 202.00 323.00 5.830000e+02 ▁▅▅▇▂
magnet_arm_z 0 1.00 306.49 326.62 -5.97000e+02 131.25 444.00 545.00 6.940000e+02 ▁▂▂▃▇
kurtosis_roll_arm 19294 0.02 -0.37 2.05 -1.81000e+00 -1.35 -0.89 -0.04 2.146000e+01 ▇▁▁▁▁
kurtosis_picth_arm 19296 0.02 -0.54 1.73 -2.08000e+00 -1.28 -1.01 -0.38 1.975000e+01 ▇▁▁▁▁
kurtosis_yaw_arm 19227 0.02 0.41 5.67 -2.10000e+00 -1.22 -0.73 0.12 5.600000e+01 ▇▁▁▁▁
skewness_roll_arm 19293 0.02 0.07 0.91 -2.54000e+00 -0.56 0.04 0.67 4.390000e+00 ▁▇▆▁▁
skewness_pitch_arm 19296 0.02 -0.06 0.83 -4.57000e+00 -0.62 -0.03 0.45 3.040000e+00 ▁▁▇▇▁
skewness_yaw_arm 19227 0.02 -0.23 1.18 -6.71000e+00 -0.74 -0.13 0.34 7.480000e+00 ▁▂▇▁▁
max_roll_arm 19216 0.02 11.24 26.93 -7.31000e+01 -0.18 4.95 26.78 8.550000e+01 ▁▂▇▃▁
max_picth_arm 19216 0.02 35.75 69.62 -1.73000e+02 -1.98 23.25 95.97 1.800000e+02 ▁▂▇▆▃
max_yaw_arm 19216 0.02 35.46 10.45 4.00000e+00 29.00 34.00 41.00 6.500000e+01 ▁▂▇▃▁
min_roll_arm 19216 0.02 -21.22 28.72 -8.91000e+01 -41.98 -22.45 0.00 6.640000e+01 ▂▆▇▂▁
min_pitch_arm 19216 0.02 -33.92 60.83 -1.80000e+02 -72.62 -33.85 0.00 1.520000e+02 ▁▆▇▂▁
min_yaw_arm 19216 0.02 14.67 9.11 1.00000e+00 8.00 13.00 19.00 3.800000e+01 ▆▇▃▂▂
amplitude_roll_arm 19216 0.02 32.45 27.39 0.00000e+00 5.43 28.45 50.96 1.195000e+02 ▇▆▃▁▁
amplitude_pitch_arm 19216 0.02 69.68 66.98 0.00000e+00 9.93 54.90 115.17 3.600000e+02 ▇▅▂▁▁
amplitude_yaw_arm 19216 0.02 20.79 12.28 0.00000e+00 13.00 22.00 28.75 5.200000e+01 ▅▅▇▃▁
roll_dumbbell 0 1.00 23.84 69.93 -1.53710e+02 -18.49 48.17 67.61 1.535500e+02 ▂▂▃▇▂
pitch_dumbbell 0 1.00 -10.78 36.99 -1.49590e+02 -40.89 -20.96 17.50 1.494000e+02 ▁▆▇▂▁
yaw_dumbbell 0 1.00 1.67 82.52 -1.50870e+02 -77.64 -3.32 79.64 1.549500e+02 ▃▇▅▅▆
kurtosis_roll_dumbbell 19221 0.02 0.45 3.19 -2.17000e+00 -0.68 -0.03 0.94 5.500000e+01 ▇▁▁▁▁
kurtosis_picth_dumbbell 19218 0.02 0.29 3.30 -2.20000e+00 -0.72 -0.13 0.58 5.563000e+01 ▇▁▁▁▁
skewness_roll_dumbbell 19220 0.02 -0.12 0.82 -7.38000e+00 -0.58 -0.08 0.40 1.960000e+00 ▁▁▁▇▆
skewness_pitch_dumbbell 19217 0.02 -0.03 0.86 -7.45000e+00 -0.53 -0.09 0.51 3.770000e+00 ▁▁▂▇▁
max_roll_dumbbell 19216 0.02 13.76 48.30 -7.01000e+01 -27.15 14.85 50.58 1.370000e+02 ▆▇▇▅▂
max_picth_dumbbell 19216 0.02 32.75 93.37 -1.12900e+02 -66.70 40.05 133.23 1.550000e+02 ▆▃▂▂▇
max_yaw_dumbbell 19221 0.02 0.45 3.19 -2.20000e+00 -0.70 0.00 0.90 5.500000e+01 ▇▁▁▁▁
min_roll_dumbbell 19216 0.02 -41.24 34.71 -1.49600e+02 -59.68 -43.55 -25.20 7.320000e+01 ▁▃▇▃▁
min_pitch_dumbbell 19216 0.02 -33.18 74.28 -1.47000e+02 -91.80 -66.15 21.20 1.209000e+02 ▆▇▅▃▃
min_yaw_dumbbell 19221 0.02 0.45 3.19 -2.20000e+00 -0.70 0.00 0.90 5.500000e+01 ▇▁▁▁▁
amplitude_roll_dumbbell 19216 0.02 55.00 54.94 0.00000e+00 14.97 35.05 81.04 2.564800e+02 ▇▂▁▁▁
amplitude_pitch_dumbbell 19216 0.02 65.93 65.23 0.00000e+00 17.06 41.72 99.54 2.735900e+02 ▇▂▂▁▁
amplitude_yaw_dumbbell 19221 0.02 0.00 0.00 0.00000e+00 0.00 0.00 0.00 0.000000e+00 ▁▁▇▁▁
total_accel_dumbbell 0 1.00 13.72 10.23 0.00000e+00 4.00 10.00 19.00 5.800000e+01 ▇▅▃▁▁
var_accel_dumbbell 19216 0.02 4.39 13.51 0.00000e+00 0.38 1.00 3.43 2.304300e+02 ▇▁▁▁▁
avg_roll_dumbbell 19216 0.02 23.86 62.90 -1.28960e+02 -12.33 48.23 64.37 1.259900e+02 ▂▂▃▇▂
stddev_roll_dumbbell 19216 0.02 20.76 24.30 0.00000e+00 4.64 12.20 26.36 1.237800e+02 ▇▂▁▁▁
var_roll_dumbbell 19216 0.02 1020.27 2262.56 0.00000e+00 21.52 148.95 694.65 1.532101e+04 ▇▁▁▁▁
avg_pitch_dumbbell 19216 0.02 -12.33 32.06 -7.07300e+01 -42.00 -19.90 13.21 9.428000e+01 ▇▇▇▂▁
stddev_pitch_dumbbell 19216 0.02 13.15 13.34 0.00000e+00 3.48 8.09 19.24 8.268000e+01 ▇▂▁▁▁
var_pitch_dumbbell 19216 0.02 350.31 673.96 0.00000e+00 12.13 65.43 370.11 6.836020e+03 ▇▁▁▁▁
avg_yaw_dumbbell 19216 0.02 0.20 78.21 -1.17950e+02 -76.70 -4.50 71.23 1.349000e+02 ▇▃▅▃▅
stddev_yaw_dumbbell 19216 0.02 16.65 17.71 0.00000e+00 3.88 10.26 24.67 1.070900e+02 ▇▂▁▁▁
var_yaw_dumbbell 19216 0.02 589.84 1244.59 0.00000e+00 15.09 105.35 608.79 1.146791e+04 ▇▁▁▁▁
gyros_dumbbell_x 0 1.00 0.16 1.51 -2.04000e+02 -0.03 0.13 0.35 2.220000e+00 ▁▁▁▁▇
gyros_dumbbell_y 0 1.00 0.05 0.61 -2.10000e+00 -0.14 0.03 0.21 5.200000e+01 ▇▁▁▁▁
gyros_dumbbell_z 0 1.00 -0.13 2.29 -2.38000e+00 -0.31 -0.13 0.03 3.170000e+02 ▇▁▁▁▁
accel_dumbbell_x 0 1.00 -28.62 67.32 -4.19000e+02 -50.00 -8.00 11.00 2.350000e+02 ▁▁▆▇▁
accel_dumbbell_y 0 1.00 52.63 80.75 -1.89000e+02 -8.00 41.50 111.00 3.150000e+02 ▁▇▇▅▁
accel_dumbbell_z 0 1.00 -38.32 109.47 -3.34000e+02 -142.00 -1.00 38.00 3.180000e+02 ▁▆▇▃▁
magnet_dumbbell_x 0 1.00 -328.48 339.72 -6.43000e+02 -535.00 -479.00 -304.00 5.920000e+02 ▇▂▁▁▂
magnet_dumbbell_y 0 1.00 220.97 326.87 -3.60000e+03 231.00 311.00 390.00 6.330000e+02 ▁▁▁▁▇
magnet_dumbbell_z 0 1.00 46.05 139.96 -2.62000e+02 -45.00 13.00 95.00 4.520000e+02 ▁▇▆▂▂
roll_forearm 0 1.00 33.83 108.04 -1.80000e+02 -0.74 21.70 140.00 1.800000e+02 ▃▂▇▂▇
pitch_forearm 0 1.00 10.71 28.15 -7.25000e+01 0.00 9.24 28.40 8.980000e+01 ▁▁▇▃▁
yaw_forearm 0 1.00 19.21 103.22 -1.80000e+02 -68.60 0.00 110.00 1.800000e+02 ▅▅▇▆▇
kurtosis_roll_forearm 19300 0.02 -0.69 2.51 -1.88000e+00 -1.40 -1.12 -0.62 4.006000e+01 ▇▁▁▁▁
kurtosis_picth_forearm 19301 0.02 0.42 4.43 -2.10000e+00 -1.38 -0.89 0.05 3.363000e+01 ▇▁▁▁▁
skewness_roll_forearm 19299 0.02 -0.01 0.73 -2.30000e+00 -0.40 0.00 0.37 5.860000e+00 ▂▇▁▁▁
skewness_pitch_forearm 19301 0.02 -0.22 1.23 -5.24000e+00 -0.88 -0.16 0.51 4.460000e+00 ▁▁▇▃▁
max_roll_forearm 19216 0.02 24.49 31.04 -6.66000e+01 0.00 26.80 45.95 8.980000e+01 ▁▁▇▇▂
max_picth_forearm 19216 0.02 81.49 95.54 -1.51000e+02 0.00 113.00 174.75 1.800000e+02 ▁▂▃▂▇
max_yaw_forearm 19300 0.02 -0.69 2.52 -1.90000e+00 -1.40 -1.10 -0.60 4.010000e+01 ▇▁▁▁▁
min_roll_forearm 19216 0.02 -0.17 22.59 -7.25000e+01 -6.07 0.00 12.07 6.210000e+01 ▁▁▇▃▁
min_pitch_forearm 19216 0.02 -57.57 110.74 -1.80000e+02 -175.00 -61.00 0.00 1.670000e+02 ▇▂▅▂▂
min_yaw_forearm 19300 0.02 -0.69 2.52 -1.90000e+00 -1.40 -1.10 -0.60 4.010000e+01 ▇▁▁▁▁
amplitude_roll_forearm 19216 0.02 24.65 25.88 0.00000e+00 1.12 17.77 39.88 1.260000e+02 ▇▃▂▁▁
amplitude_pitch_forearm 19216 0.02 139.06 147.86 0.00000e+00 2.00 83.70 350.00 3.600000e+02 ▇▃▁▁▅
amplitude_yaw_forearm 19300 0.02 0.00 0.00 0.00000e+00 0.00 0.00 0.00 0.000000e+00 ▁▁▇▁▁
total_accel_forearm 0 1.00 34.72 10.06 0.00000e+00 29.00 36.00 41.00 1.080000e+02 ▁▇▂▁▁
var_accel_forearm 19216 0.02 33.50 33.95 0.00000e+00 6.76 21.16 51.24 1.726100e+02 ▇▃▂▁▁
avg_roll_forearm 19216 0.02 33.17 79.52 -1.77230e+02 -0.91 11.17 107.13 1.772600e+02 ▁▂▇▃▅
stddev_roll_forearm 19216 0.02 41.99 59.33 0.00000e+00 0.43 8.03 85.37 1.791700e+02 ▇▁▁▁▂
var_roll_forearm 19216 0.02 5274.10 9177.18 0.00000e+00 0.18 64.48 7289.08 3.210224e+04 ▇▁▁▁▁
avg_pitch_forearm 19216 0.02 11.80 24.83 -6.81700e+01 0.00 12.02 28.48 7.209000e+01 ▁▁▇▆▁
stddev_pitch_forearm 19216 0.02 7.98 8.73 0.00000e+00 0.34 5.52 12.87 4.775000e+01 ▇▃▁▁▁
var_pitch_forearm 19216 0.02 139.59 266.49 0.00000e+00 0.11 30.43 165.53 2.279620e+03 ▇▁▁▁▁
avg_yaw_forearm 19216 0.02 18.00 77.56 -1.55060e+02 -26.26 0.00 85.79 1.692400e+02 ▂▃▇▆▃
stddev_yaw_forearm 19216 0.02 44.85 51.33 0.00000e+00 0.52 24.74 85.82 1.975100e+02 ▇▂▂▂▁
var_yaw_forearm 19216 0.02 4639.85 7284.97 0.00000e+00 0.27 612.21 7368.41 3.900933e+04 ▇▂▁▁▁
gyros_forearm_x 0 1.00 0.16 0.65 -2.20000e+01 -0.22 0.05 0.56 3.970000e+00 ▁▁▁▁▇
gyros_forearm_y 0 1.00 0.08 3.10 -7.02000e+00 -1.46 0.03 1.62 3.110000e+02 ▇▁▁▁▁
gyros_forearm_z 0 1.00 0.15 1.75 -8.09000e+00 -0.18 0.08 0.49 2.310000e+02 ▇▁▁▁▁
accel_forearm_x 0 1.00 -61.65 180.59 -4.98000e+02 -178.00 -57.00 76.00 4.770000e+02 ▂▆▇▅▁
accel_forearm_y 0 1.00 163.66 200.13 -6.32000e+02 57.00 201.00 312.00 9.230000e+02 ▁▂▇▅▁
accel_forearm_z 0 1.00 -55.29 138.40 -4.46000e+02 -182.00 -39.00 26.00 2.910000e+02 ▁▇▅▅▃
magnet_forearm_x 0 1.00 -312.58 346.96 -1.28000e+03 -616.00 -378.00 -73.00 6.720000e+02 ▁▇▇▅▁
magnet_forearm_y 0 1.00 380.12 509.37 -8.96000e+02 2.00 591.00 737.00 1.480000e+03 ▂▂▂▇▁
magnet_forearm_z 0 1.00 393.61 369.27 -9.73000e+02 191.00 511.00 653.00 1.090000e+03 ▁▁▂▇▃

Exploratory Data Analisys

From the data tab we can see that all the logical variables only have NAs and must be deleted to improve the model. Also, we can see that some features has a very low complete rate, so they have a lot of NAs. I check that with a plot

# visualizing NAs
training %>% 
  is.na() %>%
  reshape2::melt() %>%
  ggplot(aes(Var2, Var1, fill = value)) + 
  geom_raster() + 
  scale_y_continuous(NULL, expand = c(0, 0)) +
  scale_fill_grey(name = "", 
                  labels = c("Present", 
                             "Missing")) +
  xlab("Observation") +
  theme(axis.text.x  = element_text(size = 7, angle = 45,
                                    hjust = 1, vjust = 1))

From the plot we can see that there are features with a complete rate of 100% and features with just few data. So the inputation methods are not useful in this case. I also checked the proportion of every category in the categorical variables and I found that the variable new_window has a very unbalanced proportion of levels (yes = 98% and no = 2%), so this variable must also be deleted in the pre-processing procedure.

training %>%
  # select only character, factor or variables
  select(where(is.character) | where(is.factor)) %>%
  
  # calculate the proportion of unique categories in every variable
  apply(., FUN = \(x) table(x) / length(x), MARGIN = 2)
## $user_name
## x
##    adelmo  carlitos   charles    eurico    jeremy     pedro 
## 0.1983488 0.1585975 0.1802059 0.1564570 0.1733768 0.1330140 
## 
## $cvtd_timestamp
## x
## 02/12/2011 13:32 02/12/2011 13:33 02/12/2011 13:34 02/12/2011 13:35 
##      0.009020487      0.067322393      0.070074406      0.051931505 
## 02/12/2011 14:56 02/12/2011 14:57 02/12/2011 14:58 02/12/2011 14:59 
##      0.011976353      0.070329222      0.069513811      0.028386505 
## 05/12/2011 11:23 05/12/2011 11:24 05/12/2011 11:25 05/12/2011 14:22 
##      0.009683009      0.076291917      0.072622567      0.013607176 
## 05/12/2011 14:23 05/12/2011 14:24 28/11/2011 14:13 28/11/2011 14:14 
##      0.069819590      0.049587198      0.042452349      0.076342880 
## 28/11/2011 14:15 30/11/2011 17:10 30/11/2011 17:11 30/11/2011 17:12 
##      0.037661808      0.044287025      0.073387015      0.055702783 
## 
## $new_window
## x
##         no        yes 
## 0.97930894 0.02069106 
## 
## $classe
## x
##         A         B         C         D         E 
## 0.2843747 0.1935073 0.1743961 0.1638977 0.1838243

Preprocessing

Based on the fact that the data has a lot of NAs the imputation methods are not useful and the features with more than 90% of NAs should be deleted to improve the modelling. So in the first step of the presprossecing process I filtered the data to get just the features that doesn’t have NAs. In the second step I deleted the new_windows and …1 variables. I applied I did it to both the training and testing dataset. To improve the speed of the modelling process I trained the model on a sample of 1000 rows selected at random from the training dataset.

features <- training[sample(nrow(training), 1000), ] %>%
  skimr::skim() %>%
  select(skim_variable, complete_rate) %>%
  # filter complete features
  filter(complete_rate == 1) %>%
  pull(skim_variable)

train_df <- training[sample(nrow(training), 1000), ] %>%
  select(all_of(features)) %>%
  # delete new_windows and ...1 because they are near zero variance and a rowname
  # variable respectively
  select(-c(new_window, ...1))

test_df <- testing %>%
  select(all_of(features[!features == 'classe'])) %>%
  select(-c(new_window, ...1))

Model

Finally I applied a random forest model because it is versatile, robust, and suitable for various classification tasks and offers some advantages such as high accuracy, robustness to outliers, and no assumption about data distribution. I selected the resampling method ‘cv’ in the trainControl function to apply cross-validation. The Mtry parameter controls how many features are available to be considered for each new split in every tree.

Train the model

fit_rf <- train(classe ~ ., data = train_df, method = 'rf', prox = TRUE,
                trControl = trainControl(method = 'cv'))

To asses the generalization performance of the model I have estimated the out of sampling error using the k-fold cross-validation method. k-fold cross-validation is a resampling method that randomly divides the training data into k groups of approximately equal size. This process results in k estimates of the generalization error. Thus, the k-fold CV estimate is computed by averaging the k test errors, providing us with an approximation of the error we might expect on unseen data. The expected out of sample error was estimated by the accuracy measure and was about 0.96.

fit_rf
## Random Forest 
## 
## 1000 samples
##   57 predictor
##    5 classes: 'A', 'B', 'C', 'D', 'E' 
## 
## No pre-processing
## Resampling: Cross-Validated (10 fold) 
## Summary of sample sizes: 900, 900, 901, 900, 900, 898, ... 
## Resampling results across tuning parameters:
## 
##   mtry  Accuracy   Kappa    
##    2    0.9540852  0.9420556
##   40    0.9770867  0.9711073
##   79    0.9741365  0.9673938
## 
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was mtry = 40.

Predicting

The result of predicting the 20 provided test cases were the following:

predict(fit_rf, test_df)
##  [1] B A B A A E D B A A B C B A E E A B B B
## Levels: A B C D E