End-to-End ML Pipeline for Predictive Maintenance

Project Overview

The objective of this project is to develop a comprehensive machine learning pipeline for predicting equipment failures using historical time series data. The pipeline encompasses all stages of machine learning, including data ingestion, preprocessing, model training, deployment, and monitoring.

Kaggle dataset

DEMO 📌

a demo model has been uploaded to streamlit :

model monitoring dashboard:

Model Performance

Currently we are using RandomForest model, based on my EDA I was able to achieve the following scores:

The RandomForest model shows the highest accuracy and F1-score, indicating it is likely the best-performing model .

• Accuracy : 93%
  
  

Installation

To get started with this project, clone the repository and install the necessary dependencies:

git clone https://github.com/typhonshambo/End-to-End-ML-Pipeline-for-Predictive-Maintenance
cd End-to-End-ML-Pipeline-for-Predictive-Maintenance
pip install -r requirements.txt

Usage

Running the Streamlit App

streamlit run streamlit_app.py

This will start a local web server where you can interact with the predictive maintenance model through a user-friendly interface.

Experiment Tracking with MLflow

To start the MLflow tracking server, run:

mlflow ui

This will start the MLflow UI on http://localhost:5000, where you can track experiments, visualize metrics, and manage model versions.

Modeling Pipeline

The project includes a robust pipeline for data processing, model training, and prediction. The main components are:

• data_preprocessing.py: Scripts for cleaning and preprocessing raw data.
• feature_engineering.py: Functions to create and transform features.
• train_model.py: Scripts to train the predictive maintenance model.
• prediction_pipeline.py: Scripts to load the model and make predictions on new data.

Data

The data directory is structured as follows:

• raw/: Contains the raw dataset files.
• processed/: Contains processed datasets used for training and testing.
• streamlit/: Contains datasets uploaded via the Streamlit app.