Skip to content

srilekhatv/AutoNexus-Showcase

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 

Repository files navigation

🚀 AutoNexus — Professional AutoML & Explainability Platform

Open in Streamlit

AutoNexus is a secure, multi-user, no-code AutoML web application built with Streamlit and powered by a Firebase backend. It enables users to create an account, upload datasets, perform a full end-to-end machine learning workflow, and save their results. The platform guides users from data cleaning and exploratory data analysis to advanced model tuning and interpretation using SHAP and LIME.


✨ Features

AutoNexus provides a comprehensive suite of tools for any data scientist, from beginner to expert.

User & Session Management

  • 🔐 Secure User Authentication: Full login, signup, and password reset functionality powered by Firebase.
  • ☁️ Cloud Storage: (Coming Soon) Each user's datasets are stored in their own secure cloud folder.

Interactive AI Assistant

  • Conversational Guidance: Get help and explanations at any step of the workflow.
  • Context-Aware: The assistant, powered by Google's Gemini models, understands the user's current dataset and progress in the app.
  • Data Interpretation: Ask questions in plain English about your data, models, and results.

Data Preparation Workflow

  • 📁 CSV Dataset Upload: Upload datasets for analysis.
  • 🧹 Data Cleaning: A suite of tools to handle duplicates, rename columns, standardize text, and more.
  • 📊 Exploratory Data Analysis (EDA): Generate interactive plots (histograms, scatter plots, heatmaps) and statistical summaries to understand the data.
  • 🔬 Advanced Preprocessing:
    • Handle missing values with various strategies (mean, median, mode).
    • Model-Based Feature Selection using Random Forest to identify the most predictive features.
    • Statistical feature selection (ANOVA, Mutual Information).
  • ⚖️ Class Imbalance Handling: Correctly apply SMOTE, Random OverSampling, or Random UnderSampling to the training data to improve model performance on skewed datasets.

Modeling & Evaluation

  • 🤖 Model Training: Train a variety of classification and regression models, including:
    • Logistic/Linear Regression, Ridge, Lasso
    • Random Forest, Decision Trees
    • XGBoost
  • ⚙️ Advanced Hyperparameter Tuning:
    • Go beyond default settings with automated tuning.
    • Choose between GridSearchCV (exhaustive) and RandomizedSearchCV (fast and efficient) to find the best parameters for your model.
  • 📈 Robust Evaluation:
    • Utilizes a proper Train/Validate/Test split for reliable performance metrics.
    • View detailed classification reports, confusion matrices, and key regression metrics (R², MAE, RMSE).

Explainable AI (XAI)

  • 🧠 Model Interpretability: Understand why your model makes its predictions.
    • SHAP: Global and local feature importance plots.
    • LIME: Explain individual predictions on a case-by-case basis.
  • 💾 Download Artifacts: Download the final, trained model (.pkl) for offline use or deployment.

🔗 Live Demo

Click here to try AutoNexus on Streamlit Cloud


📸 Screenshots

🔹 Secure Login Page

Login

🔹 AI Assistant

AI

🔹 EDA Summary

EDA

🔹 Pre-Processing

Pre-Processing

🔹 Model Training

Model Training

🔹 LIME Explanation

LIME

🛠️ Technology Stack

  • Frontend: Streamlit
  • Backend Services: Firebase (Authentication, Cloud Storage), FastAPI (Planned)
  • Data Science: Pandas, Scikit-learn, Imbalanced-learn
  • Modeling: XGBoost, InterpretML
  • Explainability: SHAP, LIME

🙋‍♀️ Author

Srilekha Tirumala Vinjamoori   MS in Information Systems | Business Analytics | UTA

📫 LinkedIn   🌐 Portfolio


📄 License

MIT License

About

Web application for Automated Machine Learning and Explainable AI.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors