Multi-Variate Time Series Forecasting Water Quality Parameter Prediction (Regression) 

Focus of this project is develop a robust Multivariate Time Series Model utilizing a Gradient Boosting Regression (GBR) architecture to predict key Water Quality Parameters (Flow Rate, Pressure, and Temperature) at various sensor points for future planning, resource management, and setting a reliable baseline for normal system behavior.

For more details, click here

Python | Time Series | Machine Learning | Gradient Boosting | scikit-learn | BigQuery | GCP | Looker | Data Pipeline

Multi-Output Classification for Classifying energy efficiency of buildings based on Heating and Cooling load

Engineered a Multi-Output Classification model to accurately forecast building heating and cooling requirements. Through the use of sophisticated features like Roof Area and Overall Height, the project was able to identify energy inefficiencies. With approximately 99% accuracy, the XGBoost model showed outstanding predictive performance and practical insights for maximizing the use of energy resources and advancing sustainability.

This model serves as a robust tool for energy efficiency planning and data-driven decision-making in building management systems.

For more details, click here

Python | Classification | Machine Learning | Energy Efficiency | Data Analysis | XGBoost | Multi-Output Classification | SVM | Desicion Tree 

Fine-Tuning BERT for Sentiment Prediction on Amazon Alexa Reviews 

Developed a BERT-based sentiment analysis model that classifies Amazon product reviews with high accuracy, achieving a 97% F1 score. Implemented advanced data processing, including emoji handling and neutral statement classification, to enhance model performance and improve sentiment prediction accuracy by 15%.

For more details, click here

Python | Sentiment Analysis | Apache Airflow | Bert |  HuggingFace | Transformers | NLP | NoSQL | GCP