Clinker Free Lime Predictive Model

• Category: Machine Learning
• Enterprise: CEMEX
• Tools: Python, Sckit-Learn

The following research aims to summerize the step-to-step process undertaken to develop a cement quality control predictive model. The following algorithms were tested and evaluated: Logistic Regression, Support Vector Machines and Random Forest Classifier. The best performing algorithm was the Random Forest Classifier, however, evaluation metrics suggested that the model's performance wasn't considered optimal for deployment and/or production usage. Descriptive analysis indicated poor quality data that limited a good performance. Suggestions and best practices were presented to the corresponding department in order to increase dataset size and quality.