How hemanth-karumanchi/daibetes_predictor_app works
Diabetes Predictor combines an offline training and evaluation workflow with a Streamlit prediction interface. Training reads the Pima dataset, creates feature and label sets, splits the data, fits logistic regression, reports metrics, and invokes visualizations. The app accepts eight measurements, loads a serialized model, predicts a binary outcome, and displays confidence. The source shows training saves to the repository-root model file, while the app loads a separate file under model/. Their relationship is therefore not assumed. Training details also follow the README; unsampled artifact contents were not inspected.
Main entry points: main.py Training Workflow · app.py Prediction & Confidence · diabetes.csv Pima Dataset · diabetes_model.pkl Loaded Model
Components 13
Model Training
- Pima Dataset
Diabetes Data/diabetes.csv - Training Workflow
main.py - Feature / Label Split
main.py - Train-Test Split
main.py - Logistic Regression
main.py - Training Artifact
diabetes_model.pkl
Evaluation & Visuals
- Metrics & Predictions
main.py - Plot Functions
visuals.py
Prediction App
- Measurement Inputs
app.py - Loaded Model
model/diabetes_model.pkl - Prediction & Confidence
app.py - Displayed Result
app.py
Other components
- App User
Connections 13
- Pima Dataset → Training Workflow: read by
- Training Workflow → Feature / Label Split: prepares
- Feature / Label Split → Train-Test Split: provides data
- Training Workflow → Train-Test Split: splits data
- Training Workflow → Logistic Regression: fits model
- Training Workflow → Metrics & Predictions: evaluates
- Training Workflow → Plot Functions: calls plots
- Training Workflow → Training Artifact: saves model
- App User → Measurement Inputs: enters measurements
- Measurement Inputs → Prediction & Confidence: submits input
- Loaded Model → Prediction & Confidence: loaded by
- Prediction & Confidence → Displayed Result: displays prediction
- Displayed Result → App User: presents result