Closed jonathanpeppers closed 2 years ago
.\train.ps1 -seconds 500
| Top 5 models explored | ------------------------------------------------------------------------------------------------------------------ | Trainer MicroAccuracy MacroAccuracy Duration #Iteration | |1 LightGbmMulti 0.8818 0.7775 21.2 1 | |2 LinearSvmOva 0.8817 0.7820 3.7 2 | |3 FastTreeOva 0.8816 0.7911 44.4 3 | |4 LinearSvmOva 0.8807 0.7888 14.8 4 | |5 LinearSvmOva 0.8800 0.7879 8.4 5 |
After these changes, the model seems to be a bit better:
************************************************************************************************************* * Metrics for Multi-class Classification model *------------------------------------------------------------------------------------------------------------ * Average MicroAccuracy: 0.885 - Standard deviation: (.008) - Confidence Interval 95%: (.008) * Average MacroAccuracy: 0.786 - Standard deviation: (.014) - Confidence Interval 95%: (.014) * Average LogLoss: .355 - Standard deviation: (.032) - Confidence Interval 95%: (.031) * Average LogLossReduction: .324 - Standard deviation: (.04) - Confidence Interval 95%: (.04) * Average Class 0 Precision: 0.898 - Standard deviation: (.012) - Confidence Interval 95%: (.012) * Average Class 1 Precision: 0.819 - Standard deviation: (.047) - Confidence Interval 95%: (.046) *************************************************************************************************************
I also added a couple hundred rows from toxic-comments.csv.
toxic-comments.csv
After these changes, the model seems to be a bit better:
I also added a couple hundred rows from
toxic-comments.csv.