How swagath18/symptom_prediction_unsupervised_ml_model works
Project Overview
• The project is centered on unsupervised learning to predict symptoms based on mold types (and vice versa) using clustering algorithms. It leverages data preprocessing steps which include both one-hot encoding (as seen in Data_cleaning_one_hot_encoding.ipynb) and non-one-hot-encoded data preparation (data_prep_no_1hot_encoding.ipynb).
• The goal is to identify patterns and correlations between mold types, symptoms, and diagnoses using large, complex, unlabeled datasets.
Main Components
a. Data Sources:
b. Data Preprocessing & Cleaning:
• Two parallel processes:
• One Notebook for data cleaning that applies one-hot encoding (Data_cleaning_one_hot_encoding.ipynb).
• One Notebook for data preparation without one-hot encoding (data_prep_no_1hot_encoding.ipynb).
• Both processes should be shown as separate blocks since they represent alternative or experimental pipelines for preprocessing.
c. Core ML Pipeline – Unsupervised Learning:
• A block representing the Clustering Algorithm (the heart of the unsupervised model).
d. Analysis & Output:
• Following the clustering, add a block that summarizes the results:
• “Correlation Analysis & Pattern Identification”
• Possibly include “Visualization & Reporting” if the notebooks include any result visualization steps.
Relationships and Flow
• Both preprocessing blocks then funnel into the “Clustering/Pattern Detection” block.
Architectural Patterns & Technologies
• Since both notebooks serve as data preprocessing steps, they indicate a modular design. Each notebook can be thought of as separate modules that feed into a single ML pipeline.
Final Recommendations
• Components (data input, multiple preprocessing flows, model, and analysis output) are well represented with clear directional flow.