NewsClassification

Using ML.NET to classify articles, based on title and description.
Data source: https://www.kaggle.com/rmisra/news-category-dataset and changed the .json to an array of objects, for easy parsing with Newtonsoft.Json

SdcaMaximumEntropy

---

•   Metrics for Multi-class Classification model - Test Data
  

*------------------------------------------------------------------------------------------------------------

•   MicroAccuracy:    0.364
  

•   MacroAccuracy:    0.262
  

•   LogLoss:          3.858
  

•   LogLossReduction: -.177
  

---

OVA (One-Versus-All) trainer with the BinaryTrainer.

Total minutes to train with 2 iterations: 7.545 mins

---

•   Metrics for Multi-class Classification model - Test Data
  

*------------------------------------------------------------------------------------------------------------

•   MicroAccuracy:    0.58
  

•   MacroAccuracy:    0.427
  

•   LogLoss:          1.569
  

•   LogLossReduction: .522
  

---

12 iterations(31 minutes to train the model)

---

•   Metrics for Multi-class Classification model - Test Data
  

*------------------------------------------------------------------------------------------------------------

•   MicroAccuracy:    0.585
  

•   MacroAccuracy:    0.439
  

•   LogLoss:          1.873
  

•   LogLossReduction: .429
  

---