N2O predictions application with Explainable AI

This repository is a part of TMLC project with end-to-end machine learning pipeline with deployment using PaaS on streamlit. Here, we have use the concept of Explainable AI in which whatever the ML application predicts, we are able to see real time input values with its feature contributions, its impact on model, behaviour and summary in detail with.

Project title: Machine learning improves predictions of an Agricultural Nitrous Oxide (N2O) emissions from intensively managed cropping systems.

• For N2O predictions, we have used ML model with Xgb regressor algorithm and for user data analysis, wo go with Auto EDA (pandas_profiling tool). Application is integrated using Hydralit package for responsive and animated navbar, theme components in deployment using streamlit.
• An idea with data is taken from a published paper in Environmental Research Research Paper (Paper published 19 January 2021) *C

---

💡 Demo

🔖 Output -

N2O.predictions.application.with.Explainable.AI.mp4

---

🔧 Installation

• Create virtual environment python -m venv VIRTUAL_ENV_NAME and activate it .\VIRTUAL_ENV_NAME\Scripts\activate.
• Install necessary library for this project from the file requirements.txt or manually install by pip.

  pip install -r requirements.txt
  

  To create project library requirements, use below command,

  pip freeze > requirements.txt
  

• Libraries that are used in this project (highlighted few)

  import pandas as pd
  import matplotlib.pyplot as plt
  import streamlit as st
  import streamlit.components.v1 as components
  import shap
  from sklearn.metrics import r2_score, mean_squared_error
  from pandas_profiling import ProfileReport
  from streamlit_pandas_profiling import st_profile_report
  import pybase64, joblib

---

• Run app.py using below command to start streamlit API

  streamlit run app.py
  

• By default, streamlit will run on port 8501 locally.

  Local URL: http://localhost:8501

  Network URL: http://192.168.0.102:8501

---

🔖 Directory Structure

    .                                               # Root directory
    ├── data                                        # Input data directory
    │   ├── Dataset.csv                             # Dataset used for Xgb model
    │   ├── sample_data.csv                         # Sample input data
    ├── Model                                       # Model directory
    │   ├── model_xgb_61_final.joblib.compressed    # Model (Xgb regressor)
    ├── apps                                        # Application directory
    │   ├── app1.py                                 # Single prediction with AI
    │   ├── app2.py                                 # Predictions on user data with AI
    │   ├── app3.py                                 # UI enhancement
    │   ├── app4.py                                 # Connect page
    ├── .streamlit                                  # Streamlit theme param directory
    │   ├── config.toml                             # Configurarion theme file
    ├── requirements.txt                            # Project requirements library with versions
    ├── README.md                                   # Project README file
    └── LICENSE                                     # Project License file

---

☁️ Live Application

Live Aplication is running on streamlit cloud platform, you can access from below.

---

📱 Connect with me

You say freak, I say unique. Don't wait for an opportunity, create it.

Let’s connect, share the ideas and feel free to ping me...