🛠️ Maintenance Mode Prediction with Decision Trees

Predicting Machine Operational States (Failure vs. Production) using Sensor & Maintenance Data
*Statistics & Data Analysis Course Project *

📌 Overview

In industrial systems, unexpected machine failures lead to costly downtime. This project leverages real-world maintenance logs to build a Decision Tree classifier that predicts whether a machine is in Failure (0) or Production (1) mode - enabling proactive maintenance strategies.

Developed as part of a Statistics and Data Analysis course, this work bridges Industrial Engineering (reliability, operations) with Computer Science (machine learning, data pipelines).

🔍 Problem Statement

Given sensor readings and operational features (e.g., Age, Pressure, Run_Hours, Sensor5), classify the machine’s current operational mode to support predictive maintenance decisions.

📊 Dataset

• Source: Maintenance_St.xlsx (353 samples, 13 features)
• Features:
  • Age, Machine_Type, Pressure, Temperature
  • Engine_Problem, Run_Hours, Maintenance
  • Sensor data (Sensor1–Sensor5)
• Target: Operation_modes (0 = Failure, 1 = Production)
• Class Distribution: ~73% Production, ~27% Failure

🧪 Methodology

1. Data Cleaning: Removed corrupted rows, handled missing values.
2. Exploratory Data Analysis (EDA):
   • Boxplots & histograms for outlier detection
   • Correlation heatmap (Sensor5 and Operation_modes show strong negative correlation: -0.81)
3. Modeling:
   • Trained a Decision Tree Classifier (sklearn)
   • Evaluated with accuracy, precision, recall, F1-score, confusion matrix
4. Optimization:
   • Hyperparameter tuning via manual refinement → 96% accuracy
   • Key parameters: max_depth=3, min_samples_leaf=4, min_samples_split=2
5. Interpretability:
   • Visualized decision rules
   • Extracted feature importance (Sensor5, Age, Sensor2 most influential)

📈 Results

Metric	Score
Base Accuracy	94.37%
Optimized Accuracy	96%
Failure Recall	98% (critical for maintenance!)
Key Features	Sensor5, Age, Sensor2

✅ Prediction for given input:
Age=55, Machine_Type=1, Pressure=140, ..., Sensor5=20 → Production (1) ✅

🚀 How to Run

1. Clone this repo: ```bash git clone https://github.com/lewisndambiri/maintenance-mode-predictor.git cd maintenance-mode-predictor

2. Install dependencies: ```bash pip install -r requirements.txt

3. Launch Jupyter and run the notebook: ```bash jupyter notebook maintenance_analysis.ipynb

Note: The dataset Maintenance_St.xlsx must be in the same directory.

🛠️ Tech Stack

• Languages: Python
• Libraries: pandas, numpy, scikit-learn, matplotlib, seaborn
• Tools: Jupyter Notebook, Excel

💡 Future Improvements

• Handle class imbalance with SMOTE or class_weight=’balanced’
• Experiment with Random Forest or XGBoost for higher robustness
• Deploy as a Flask API for real-time predictions
• Add SHAP values for deeper interpretability

📄 License MIT License — see LICENSE for details.

Built by Lewis NDAMBIRI