QuickStart - minimal configuration guide

QuickStart - minimal configuration guide#

You can copy one of the tutorial notebooks found in this documentation and customize it for your project. The customization includes setting serveral parameters depending the type of the analysis.

🚀 Easy Configuration with Streamlit App#

NEW: Use our interactive configuration interface!

Instead of manually setting parameters, you can use our Streamlit-based configuration interface:

The Streamlit app provides:

  • ✅ Interactive parameter configuration with validation

  • ✅ Real-time preview of your settings

  • ✅ Export ready-to-run Python scripts

  • ✅ Built-in help and documentation

  • ✅ No manual coding required

Simply configure your pipeline through the web interface and download the generated Python script!

Manual Configuration (Alternative Method)#

General Settings (Mandatory for All Tasks)#

  1. Specify Main Directory:

    • main_folder_name = 'results_<ProjectName>'

Binary Classification (Mandatory Parameters)#

  1. Outcome Variable:

    • outcome_var = "target"

you must also specify the names of categorical variables (features) if available in your dataset, like cat_features = ["race", "gender"]. Also if you need some variables not to be included like columns_to_drop = ["irrelevant feature X", "irrelevant feature Y"], and also to make sure continuous variables are also defined like continuous_features = ["age","weight"].

  1. Class Labels:

    • Positive Class: class_1 = "malignant"

    • Negative Class: class_0 = "benign"

    also replace 0 and 1 here based on your dataset class_label_dict = {0:class_0, 1:class_1}, maybe in your dataset they’re like -1 and 1 instead.

  2. Data Split:

    • Enable/Disable: data_split = True

  3. Training Size:

    • If data_split = True:

      • train_size_perc = 0.8

  4. Already Split Data (If Applicable):

    • already_split = False

    • If True, specify subsets:

      testset = mydata[mydata['subset'] == 'Test']
mydata = mydata[mydata['subset'] == 'Train']

  5. Classification Models:

    • Specify at least two models:

Survival Analysis (Mandatory Parameters)#

  1. Enable Survival Analysis:

    • survival_analysis = True

  2. Time-to-Event Column:

    • time_to_event_column = "max_time_difference_days"

Regression (Mandatory Parameters)#

  1. Enable Regression Analysis:

    • regression_analysis = True

  2. Outcome Variable:

    • regression_outcome = "regression_outcome_var"

Resource Configuration#

  1. CPU/GPU Settings:

    • Enable GPU (if available): GPU_avail = True

    • CPUs for Training: n_cpu_model_training = 20

    • CPUs for Hyperparameter Tuning: n_cpu_for_tuning = 20

Additional Notes#

  • Recommended: Use the Streamlit configuration app for easier setup and automatic parameter validation

  • Each task (classification, survival, regression) has specific mandatory parameters. Only include the settings relevant to the task you are running.

  • Omitting any of the above parameters will result in errors or crashes during execution.

  • For further details, refer to the full pipeline manual.