1.1 Machine Learning Techniques

The solution uses in-database OML4SQL, which is available by default with database version 19c and above.

1. Multi-Class Classification
   1. The machine learning framework builds individual model, for each of the following industry accepted algorithms,
      1. Decision Tree
      2. Naive Bayes
      3. Random Forest
      4. Support Vector Machines
      5. Neural Network
      6. Generalized Linear Models
   2. When we compare models, we select the one that best represents the underlying data.
   3. Model Metrics used to compare models,
      1. AUC(Area Under the Curve) – The model’s robustness is measured by this value, which is between 0 and 1. The closure the value is to 1, the better the model.
      2. Accuracy – Measured as a function of predicted vis-à-vis actual in the testing dataset.
   4. The important attributes, their individual weights and constant value used in building the model are stored in the database.
2. Regression
   1. The machine learning framework builds individual model, for each of the following industry accepted algorithms,
      1. Support Vector Machines
      2. Neural Network
      3. Generalized Linear Models
   2. The models are compared, and the best model is selected which best represents the underlying data.
   3. Model Metrics used to compare models,
      1. RMSE(Root Mean Square Error)
   4. The important attributes, their individual weights and constant value used in building the model are stored in the database.

   Note:

   Overall, there are one multi-classification model and three regression models for this solution