Pump_failure_training

Predictive Maintenance: Failure Classifier - Workshop 🧪

Overview 📋

This hands-on workshop guides participants through building and evaluating various classification models to predict equipment failures. Participants will learn by working through a series of Jupyter notebooks, from data loading and preparation to model training and evaluation, and finally deploying a model with a Streamlit application.

What You’ll Learn 🧠

• Loading, inspecting, and cleaning real-world data.
• Performing data transformations to prepare data for modeling.
• Establishing a baseline model for comparison.
• Implementing and fine-tuning tree-based classification models:
  • Random Forest
  • XGBoost
  • LightGBM (LGBM)
• Using AutoML with PyCaret to find the best model.
• Evaluating model performance using appropriate metrics.
• Understanding the workflow of a machine learning project from data to model.
• Deploying a model with a Streamlit application.

Getting Started 🛠️

✅ Recommended Platform: Google Colab. Google Colab provides a free, interactive environment that’s ideal for this workshop. No local installation is required!

What You Need:

• A Google account.
• A reliable internet connection.

Running the Notebooks in Colab:

1. Access the Notebooks:
   • Open the main GitHub repository page for this workshop.
   • Navigate to the soluciones/ directory.
2. Open in Colab:
   • Click on a notebook file (e.g., 01_load_and_clean_data.ipynb).
   • Look for an “Open in Colab” badge/button at the top of the notebook preview on GitHub. Click it.
   • Alternatively, if the badge isn’t available:
     • On the GitHub notebook page, click the “Raw” button. Copy the URL from your browser’s address bar.
     • Open Google Colab (https://colab.research.google.com/).
     • Select File > Open notebook.
     • Choose the “GitHub” tab, paste the URL, and press Enter.
3. Install Dependencies (in Colab):
   • Once a notebook is open in Colab, the first code cell in many notebooks will be for installing necessary libraries from the requirements.txt file.
   • Run this cell by pressing Shift+Enter or clicking the play button.

Running the Streamlit Application Locally:

1. Clone the repository:

   git clone <repository-url>
   cd <repository-folder>
   

2. Create a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`
   

3. Install dependencies:

   pip install -r requirements.txt
   

4. Run the app:

   streamlit run main.py
   

   Your web browser should open with the application running.

Deploying to Streamlit Cloud:

For a live, shareable version of your application, you can deploy it to Streamlit Cloud for free.

1. Push your code to a GitHub repository.
2. Sign up for Streamlit Cloud: Go to share.streamlit.io and sign up with your GitHub account.
3. Deploy the app:
   • Click on “New app”.
   • Select your repository and the branch where your code is.
   • Ensure the “Main file path” is set to main.py.
   • Click “Deploy!”.

Workshop Sessions 📚

Each session corresponds to a Jupyter notebook in the soluciones/ directory.

Session	Notebook	Topic
1	01_load_and_clean_data.ipynb	Loading, Inspecting & Cleaning Data
2	02_transform_data.ipynb	Data Transformation
3	03_baseline_model.ipynb	Creating a Baseline Model
4	04_random_forest_model.ipynb	Random Forest Classification
5	05_xgboost_model.ipynb	XGBoost Classification
6	06_light_gbm_model.ipynb	LightGBM (LGBM) Classification
7	07_automl_pycaret.ipynb	AutoML with PyCaret
(Optional)	main.py (Streamlit app)	Deploying a Model with Streamlit (Demo)

Learning Outcomes 🎯

By the end of this workshop, you’ll be able to:

• Confidently load, clean, and prepare data for machine learning tasks.
• Apply various data transformation techniques.
• Build, train, and evaluate several industry-standard classification models in Python.
• Understand the differences and trade-offs between Random Forest, XGBoost, and LightGBM.
• Use AutoML to automate the model selection process.
• Follow a structured approach to solving classification problems with machine learning.
• Deploy a machine learning model as a web application using Streamlit.

Repository Structure 📁

.
├── figuras/                     # Contains images used in the Streamlit application
│   └── icone.png
│   └── maintenance.jpg
├── soluciones/                  # Workshop notebooks: from data processing to modeling
│   ├── 01_load_and_clean_data.ipynb
│   ├── 02_transform_data.ipynb
│   ├── 03_baseline_model.ipynb
│   ├── 04_random_forest_model.ipynb
│   ├── 05_xgboost_model.ipynb
│   ├── 06_light_gbm_model.ipynb
│   └── 07_automl_pycaret.ipynb
├── data/                        # Contains datasets for the workshop
│   ├── raw_data.csv
│   ├── clean_data.csv
│   └── transformed_data.csv
├── main.py                      # Example Streamlit application script for model deployment
├── model/                       # Contains a pre-trained example model and preprocessing objects
│   ├── final_model.joblib
│   ├── label_encoder.pkl
│   └── preprocessor_pipeline.pkl
├── requirements.txt             # Lists Python dependencies for local setup / Colab
└── utils.py                     # Utility functions for metrics and preprocessing

Prerequisites 📾

• Basic Python skills: Familiarity with data types, loops, functions, and basic syntax.
• Some knowledge of basic machine learning concepts: Understanding of terms like features, target, training, testing, and model evaluation.
• Familiarity with pandas and scikit-learn (optional): Helpful, but not strictly required. The workshop will guide you through their usage.

Additional Resources (Optional) 📚

• Python Data Science Handbook by Jake VanderPlas
• Scikit-learn User Guide
• XGBoost Documentation
• LightGBM Documentation
• PyCaret Documentation
• Streamlit Documentation