Quick start
YOUR FIRST ANALYSIS
Prepare your data, configure an analysis, and export a Python script. Start with the interactive app or adapt a worked tutorial.
Open the configuration app Browse tutorials →Choose your dataset and outcome.
Select models and analysis settings.
Run the script in your analysis environment.
Configure with the app
The MAIT configuration app guides you through data settings, model selection, feature engineering, training, and outputs. It prepares a Python script; run that script in your local MAIT analysis environment.
- Set up your data. Upload a CSV to preview its columns, or enter a file path. Choose the outcome and categorical features, then exclude any columns you do not want to use.
- Choose your settings. Work through the sections in the sidebar. Select models, review the data split, and set the training and output options.
- Export your analysis. Open Finalize and Export, review the summary, and choose Export Python script, then Download Python Script. Save it in your MAIT directory and follow the setup instructions to run it.
For an example, choose File Path → Use Demo Data. Set the outcome to heart_disease, select gender and smoking_status as categorical features, and exclude patient_id.
Run the app locally
From a copy of the MAIT repository, install the interface requirements in a Python environment, then launch it:
python -m pip install -r streamlit_requirements.txt
python -m streamlit run streamlit_app.py
These requirements install the configuration interface. Use the analysis setup instructions for running the exported pipeline.
Configure a tutorial manually
Prefer to work in a notebook? Copy a worked tutorial and edit its configuration block. The examples below highlight settings to review; retain the tutorial’s other settings and imports.
Data and output
Use column names exactly as they appear in your dataset. Separate categorical and continuous predictors, and exclude identifiers or variables that should not enter the model.
main_folder_name = "results_my_project"
outcome_var = "target"
cat_features = ["race", "gender"]
continuous_features = ["age", "weight"]
columns_to_drop = ["patient_id"]
Binary classification
Define the negative and positive class labels and map them to the values used in your outcome column. In this example, the data are encoded as 0 and 1; adapt the keys if your encoding differs.
class_0 = "benign"
class_1 = "malignant"
class_label_dict = {0: class_0, 1: class_1}
Select the classification models in the tutorial’s model configuration. Start with a small selection to check the workflow before expanding your comparison.
Training and test data
For an 80% training / 20% test split:
data_split = True
train_size_perc = 0.8
already_split = False
For a predefined split, enable already_split and follow the manual’s data-splitting instructions. For repeated observations, review patient-level splitting before training.
Survival analysis and regression
Use a tutorial appropriate to your task and review these additional settings:
| Task | Settings to review | What they specify |
|---|---|---|
| Survival analysis | survival_analysis = True |
Enable the survival workflow. |
time_to_event_column = "max_time_difference_days" |
Name of the time-to-event column; also check the tutorial’s event/outcome coding. | |
| Regression | regression_analysis = True |
Enable the regression workflow. |
regression_outcome = "regression_outcome_var" |
Name of the continuous outcome column. |
Computing resources
Set CPU counts to suit your computer. Enable GPU use only when your hardware and the selected models support it.
GPU_avail = False
n_cpu_model_training = 1
n_cpu_for_tuning = 1