DiabetesCare-AI

personalised diabetes prediction along with a Q & A Chatbot

File Structure

• finalapp.py: The main script for running the Streamlit web app.
• requirements.txt: Lists all the dependencies required for the project.
• wowmodel2.pkl: The pre-trained machine learning model.
• diabetes_prediction_dataset.csv: The dataset used for predictions and visualizations.

Preprocessing, Model Building and Hyperparameter Tuning

• Conducted one-hot encoding categorical features
• Implemented Random Forest with GridSearchCV for Hyperparameter Tuning

Results:

• Selected best parameters after cross-validation.
• Evaluated model performance on the test set.
• Achieved high accuracy 0.94 on the test set.

Web App Components

1. User Authentication

   Users can log in using their name. The session state is used to manage user accounts and their prediction history

2. Data Input

   The sidebar allows users to input various health indicators such as gender, age, hypertension, heart disease, smoking history, height, weight, HbA1c level, and blood glucose level.

   BMI is calculated automatically based on the height and weight inputs.

3. Prediction

   The predict_button triggers the prediction function which uses the pre-trained model to predict the likelihood of diabetes. The prediction result is displayed to the user.

   The pre-trained achieving an accuracy of 94% on a 100000 large dataset was loaded as a .pkl file into the streamlit code.

4. Visualisation

   Seaborn and Matplotlib enables the users to examine the relationship between the features. Multiple graphs integrated to provide a comprehensive overview of the input data and the diabetes_prediction_dataset.csv.

   If diabetic the plot on the graph is a unique circle with a shade of red, else the circle plotted is dark shade of blue.

5. Suggestions

   Personalised lifestyle and dietary suggestions, including helpful resources of hospitals in India are provided by the integrtaion of gemini-1.5-flash LLM model.

   Sutable safety_settings and temperature was configured along with the nucleus sampling of the top_k and top_p temperature

   Kept maximum output of 4096 tokens at a time for the assistance.

6. Q & A Chatbot

   The website hosts a Q & A Chatbot to answer queries arising by patients. The history of queries entered by the user are saved and displayed in the end. The chatbot leverages the use of gemini-1.5-flash LLM Model.