• Vision Transformer (ViT): Applied to computer vision tasks for classifying waste image items as either recyclable or non-recyclable.
  • Natural Language Processing (NLP): Utilized BERT and BERTweet (a pre-trained language model for English tweets) to evaluate essay scoring for English language learners in grades 8-12.
  • Convolutional Neural Networks (CNNs): Developed for image classification and recognition tasks, including:
    • Classifying IDC breast cancer histopathology images as cancerous or non-cancerous.
    • Predicting flight departure delays.
  • Transfer Learning with CNNs: Leveraged pre-trained models such as VGG16, VGG19, ResNet50, ResNet152, DenseNet201, Xception, InceptionV3, EfficientNetB7, and MobileNetV for improved training efficiency and performance mainly for image classification tasks.
  • Long Short-Term Memory (LSTM): Applied to sentiment analysis in the energy industry to forecast the performance of energy-related mutual fund benchmarks.
  • Multi-Layer Perceptrons (MLPs): Applied to various classification and regression problems.
  • XGBoost: Effective for structured data with high interpretability, especially on large datasets.
  • Decision Trees & Random Forests: Used for robust predictive modeling with feature importance analysis.
  • Regression Techniques:
    • Linear & Logistic Regression: Applied to basic predictive tasks.
    • General Linear Model (GLM): Utilized for complex data relationships.
  • Clustering Techniques:
    • K-Means Clustering: Used for unsupervised data segmentation.
  • Time Series Modeling:
    • Seasonal ARIMA, STL Decomposition, Exponential Smoothing, and Prophet: Employed for forecasting and trend analysis on energy commodities.