Failure Prediction Model

Maximise Energy Company's Profits through Data Science

Objective and Focus

Machine Learning techniques to be used to predict part failures before they can occur

Data Profile

• Weather data | For each site
• Site and Sensor data
• Revenue data | For each site
• Repair costs | For each part repaired in the past (Available only in training data)

Approach

Requirements:

• Python 3.7.0.
• Pandas
• sklearn
• matplotlib.pyplot
• seaborn
• RandomForestRegressor
• StandardScaler

Model Results

Prediction to fail results

• Model trained using: Sensor data,Weather data
• Model Type: Logistic Regression, Random Forest
• Accuracy: 97.8%
• Macro F1 Score: 69.3%