Project Showcase

Highlight of my featured works

Air Quality Forecasting

Time Series 2025

A multivariate time series project forecasting air quality index (AT) using deep learning. Includes decomposition, scaling, and LSTM/BiLSTM training with evaluation.

Air Quality Trends Training Loss Curve Prediction vs Actual Results

🎯 What I Learned

  • Performed seasonal decomposition and PACF analysis
  • Applied RobustScaler and MinMaxScaler preprocessing
  • Built LSTM and BiLSTM models for forecasting

🛠 Tech Stack

  • Python, Pandas, NumPy
  • Matplotlib, Seaborn
  • Statsmodels, Scipy
  • Scikit-learn
  • TensorFlow / Keras

📌 Evaluation / Next Step

BiLSTM outperformed the baseline and LSTM models by capturing bidirectional dependencies, achieving lower MSE/MAE and higher R². However, sudden external factors such as pollution spikes remain a limitation. Next steps: integrate weather/traffic data, hybrid ARIMA–LSTM models, and deployment for real-time forecasting.

View on GitHub