2025-07-10

ATLAS — Artificial Taxonomic Learning and Analysis System

AI-driven biological identification using environmental DNA (eDNA) datasets, achieving 89% classification accuracy.

The Problem

Traditional biodiversity tracking requires manual collection and visual taxonomy of organisms, which is slow, expensive, and error-prone.

The Solution

ATLAS uses machine learning classifiers to analyze environmental DNA (eDNA) sequences, automating biodiversity taxonomy.

Key Capabilities

Robust ETL preprocessing pipeline to clean raw biological gene sequence datasets.
Supervised classification models (Random Forest, SVM) trained to identify species signatures.
Interactive metrics reporting for accuracy, precision, and recall ratios.

Technology Stack

PythonNumPyPandasScikit-learnMatplotlib

Deep Taxonomy Classification

ATLAS implements a data processing workflow from raw genomic tables to species labels:

[Raw eDNA CSV] -> [Pandas Preprocessing] -> [TF-IDF/K-mer Feature Extraction] 
                      |
                      v
             [Scikit-learn Classifier] -> [89% Accuracy Tag Report]

Performance & Evaluation

Model Accuracy: Attained 89% validation accuracy on test sequence datasets.
Orchestration: Used data preprocessing pipelines that can easily scale to handle thousands of concurrent genomic sequences.

Lessons Learned & Key Takeaways

This research project deepened my data science skills: handling highly unbalanced datasets, feature engineering on genomic strings, and selecting optimal hyperparameters for maximum SVM classifier accuracy.