Welcome to the project for the course-" Machine Learning for Energy System Applications". Workflow on an energy systems application of risk assessment – Data pre-processing, standardization, normalization. – Feature selection – Model selection and training

The data pre-processing consists of three sub tasks:

• Feature removal based on variance
• Create 'reduction' dataset for both regression and classification task
• Create 'statistical' dataset for both regression and classification task

The following tasks are performed under this:

• Compare the following models performance based on default parameters
  • Lasso regression
  • Ridge regression
  • SVR- linear
  • SVR- polynomial
  • SVR- gaussian
  • Deep neural network
• Grid search hyper parameter for the best chosen model (Neural network model)
  • This task is done for each type of dataset
• Classify risk state based on risk factor predicted by the regression model

The following tasks are performed under this: Calling the classification pipeline function does everything

• If Choose model is said to True:

  • When calling the classification pipeline function:
  • specify the dataset type to be read
  • specify the chosen model
  • Choose and fine tune any of the models:
    • SVC- Linear
    • SVC- Polynomial
    • SVC-RBF
    • Random Forests Classifier
    • Extreme Trees Classifier
    • MLP Neural Network
  • Get the validation accuracy, and test accuracy, per class precision, recall and f1 score for the fine-tuned model, also get the trained model

• If Choose model is said to False:

  • When calling the classification pipeline function:
  • specify the hyperparameters dictionary with the final chosen hyperparameters
  • specify the dataset type to be read
  • specify the chosen model
  • The final model of type 'x' is chosen and the test results are printed, the trained model is returned

• Download everything as submitted into a single folder
• Go to 'main.py' file
• Depending on the task un comment and run the pipeline functions.