This project is centered around developing a smart alarm system that leverages machine learning to predict the optimal time to ring the alarm based on the user's sleep patterns and snooze behavior.
- Using the Google Fitness API, the project retrieves sleep data from smart wearables. By analyzing this data, including bedtime, sleep cycles(N1,N2,N3,REM), and waking time, an ML model is trained to predict the user's snooze behavior. The model estimates the potential number of snoozes the user might take after the initial alarm rings.
- Leveraging the predicted snooze count and considering the user's desired wake-up time, the system calculates the optimal time to initially set the alarm. This calculation takes into account the snooze duration, ensuring that the user can still get the desired amount of sleep despite potential snoozes.
- Download the dataset : Save the dataset ('kk.csv') locally
- Train the ML model : Open the file 'aitrain.ipynb' . Ensure you have the required libraries installed:
numpy,pandas,matplotlib,tensorflow,scikit-learn. Run the script to train the machine learning model using the dataset - Prepare the pickle model : Open the file 'ai.py' and run the script to create the model 'finalized_model.sav' for snooze count prediction
- Start the Web Interface :
- Google Calendar Integration :Open the 'cal.py' file and follow the instructions (commented out) to integrate with Google Calendar. Run the script to retrieve sleep data using the Google Fitness API
- Travel Time Calculation : Run the 'map.py' file to calculate the travel time between two locations using the Open Source Routing Machine API
- Optimal Alarm Time Calculation : Open the 'main.py' file and run the script to calculate the optimal alarm time