Machine Learning Model for Detecting Faults in Variable Refrigerant Flow (VRF) Systems

This study focuses on building a machine learning model to forecast maintenance requirements for Variable Refrigerant Flow (VRF) systems by examining their operational data. A detailed dataset, obtained from Kaggle.com, includes essential parameters such as temperature, pressure, compressor behaviour and energy usage along with maintenance and failure logs. The data is thoroughly preprocessed to enhance its accuracy and dependability, followed by exploratory analysis to identify key patterns and correlations. Predictive models like Random Forest (RF), Support Vector Machine (SVM) and K-Nearest Neighbour (KNN) are evaluated to determine which offers better classification performance. These models are then trained and validated using Accuracy score, F1-score, Precision and Recall respectively.