Skip to content

Latest commit

 

History

History
83 lines (74 loc) · 3.58 KB

README.md

File metadata and controls

83 lines (74 loc) · 3.58 KB

Automl comparison

This is code for comparison of automatic machine learning libraries:

Datasets used for comparison

Dataset Id Name Rows Columns
3 kr-vs-kp 3196 36
24 mushroom 8124 22
31 credit-g 1000 20
38 sick 3772 29
44 spambase 4601 57
179 adult 48842 14
715 fri_c3_1000_25 1000 25
718 fri_c4_1000_100 1000 100
720 abalone 4177 8
722 pol 15000 48
723 fri_c4_1000_25 1000 25
727 2dplanes 40768 10
728 analcatdata_supreme 4052 7
734 ailerons 13750 40
735 cpu_small 8192 12
737 space_ga 3107 6
740 fri_c3_1000_10 1000 10
741 rmftsa_sleepdata 1024 2
819 delta_elevators 9517 6
821 house_16H 22784 16
822 cal_housing 20640 8
823 houses 20640 8
833 bank32nh 8192 32
837 fri_c1_1000_50 1000 50
843 house_8L 22784 8
845 fri_c0_1000_10 1000 10
846 elevators 16599 18
847 wind 6574 14

To download datasets you need to register on openML and set OPENML_KEY in your environment.

Methodology

  1. Each dataset was divided into train and test set (70%/30%).
  2. The autoML package was trained on train set. There was 1 hour limit for training.
  3. Final autoML model was used to compute predictions on test set (samples not used for training).
  4. The logloss was used to asses model performance (the lower the better).
  5. The process was repeated 10 times (with different seeds), results are average over 10 repeats.

Results

Dataset Id Auto-sklearn H2O MLJAR
179 0.4977899919 0.3152976149 0.3049036708
24 0.0008299585 0.0064581712 0.000003843
3 0.1971600449 0.0259192684 0.0188968126
31 0.5083364979 0.5619242106 0.4939436971
38 0.1654747739 0.045179262 0.0389534345
44 0.3838450926 0.1360053817 0.125079404
715 0.2818374134 0.236469178 0.2068510073
718 0.2638171184 0.2789765963 0.2419061124
720 0.4998641475 0.4516830437 0.4309945916
722 0.3908241383 0.0322915583 0.0298104695
723 0.3293591298 0.2739778619 0.2500752244
727 0.3978124619 0.1595664723 0.1495902549
728 0.0927113633 0.0331800086 0.0198811926
734 0.5432702419 0.2738200987 0.2548240677
735 0.4268106101 0.1811306173 0.1639352045
737 0.4824914304 0.330549118 0.367133044
740 0.3142420934 0.2254379351 0.2149655921
741 0.5729526032 0.5671571325 0.5247308762
819 0.4847074932 0.2953733834 0.2865255395
821 0.5601634059 0.2545560683 0.241073802
822 0.5179014444 0.2326348019 0.2373530782
823 0.4203668177 0.0664409033 0.0403094539
833 0.442981634 0.3944518914 0.3769150535
837 0.3427939126 0.2430881733 0.2064752939
843 0.4363679407 0.2670714458 0.2531006498
845 0.4559840276 0.3148631409 0.2550298912
846 0.5457531577 0.2467550186 0.2401925153
847 0.3860761231 0.3245902333 0.3049773203