This project explores a predictive maintenance dataset and compares several machine learning models for detecting likely machine failures before they happen.
The notebook trains and evaluates classifiers on the AI4I 2020 Predictive Maintenance dataset to predict whether a machine will fail based on sensor and operating conditions.
- Logistic Regression
- Random Forest
- Support Vector Machine (SVM)
The analysis uses the AI4I 2020 Predictive Maintenance Dataset, which contains sensor readings and operational metadata for machine health monitoring.
The comparison shows that the Random Forest model achieved the strongest overall balance between accuracy and F1 score for this task.
- Predictive_Maintenance_Code.ipynb — notebook with data loading, preprocessing, model training, and evaluation
- data/dataset.csv — local copy of the dataset used for the analysis
- images/ — generated plots and performance visuals
- requirements.txt — Python dependencies for reproducing the notebook
- Create and activate a virtual environment.
- Install dependencies:
pip install -r requirements.txt
- Open the notebook in Jupyter and run the cells.
This project gave me practical experience with supervised machine learning by building and comparing multiple classification models.
Through this project I gained experience with:
- Supervised machine learning workflows
- Classification problems
- Logistic Regression
- Random Forest
- Support Vector Machines (SVM)
- Data preprocessing and feature scaling
- Model evaluation using Accuracy, Precision, Recall and F1 Score
- Comparing algorithms to identify the best-performing model
- Understanding how dataset quality affects prediction reliability
The biggest takeaway was that a machine learning model is only as good as the data it is trained on. Poor-quality or unrepresentative datasets lead to unreliable predictions, regardless of how advanced the algorithm is.