Audiogram Predict

This is a Machine Learning project to predict new frequencies values

Installation

To install required packages:

./scripts/install.sh

Execution

To execute the application run:

poetry run streamlit run src/main.py

Versioning

Experiments: Conda Environment (conda_environment.yml)

notebook/

Script: Poetry (pyproject.toml)

src/

Data: DVC connected on AWS S3 Bucket (.dvc/)

dataset/
model/

Files tracked by Git and DVC:

DVC connection to AWS S3:

Modeling

The application is served by KNN models to predict the best value. The machine learning pipeline follow the best correlation from the input combination. E.g.: For the combination '2k' - '4k' - '6k' the most correlated is the '3k' and the next prediction.

Machine Learning Pipelines:

• Comb: 2k-4k-6k -> 3k

• Comb: 2k-4k-6k-1k -> 3k
  Comb: 2k-4k-6k-3k -> 1k
  Comb: 2k-4k-6k-8k -> 3k
  Comb: 2k-4k-6k-500k -> 3k

• Comb: 2k-4k-6k-8k-3k -> 1k
  Comb: 2k-4k-6k-1k-3k -> 500k
  Comb: 2k-4k-6k-500k-3k -> 1k

• Comb: 2k-4k-6k-500k-3k-1k -> 8k
  Comb: 2k-4k-6k-8k-3k-1k -> 500k

Experiments

In order to run the notebooks run: Create a new conda environment and activate:

conda create -n cond_audio_env python=3.10
conda activate cond_audio_env

Install the required dependencies:

conda env update --file conda_environment.yml

Open Jupyter

jupyter-notebook