BARRY
  • Machine Learning Case Studies in Finance

    • Boosted portfolio profit by 30% through a sentiment-based trading strategy using BERT variants and GPT-2 LLMs
    • Improved stock prediction by 30% over the baseline random forest model using fine-tuned logistic regression models
    • Developed a multi-indicator-based hierarchical trading strategy, leading to over 150% profit gain during testing

    Key words: trading strategy, LLM, NLP, PCA, supervised learning

  • Production Assessment of E. coli Model with Synthetic CO2 Fixation Pathways

    • Designed a novel metabolic model by integrating synthetic reductive glycine pathway (rGlyP) and synthetic acetyl-CoA pathway (SACA) into the E. coli iML1515 model
    • Identified 15 key reactions for integrating 2 synthetic CO2 fixation pathways into the model from over 10 literatures
    • Optimized and evaluated the model using Cobra through cplex linear programming, flux analysis, knock-out analysis, growth coupling analysis

    Key words: rGlyP, SACA, metabolic modeling, linear programming, cplex, bioproduction, carbon footprint, sustainable energy

  • Curriculum Design via Hierarchical/K-Means/DBSCAN Clustering

    • Evaluated 3 clustering models and optimized K-Means and DBSCAN clustering models using the elbow method and grid search
    • Experimented with sub-clustering on input features by by applying word embedding through transformer models
    • Tested generative AI for summarizing the final results

    Key words: categorical variable, feature engineering, hierarchical clustering, K-Means, DBSCAN, unsupervised learning, generative AI

  • Compensation Classification via Multi-class Ordinal Logistic Regression

    • Implemented multi-class ordinal logistic regression based on pseudo steps
    • Engineered features through exploratory analysis, feature encoding, and feature selection
    • Enhanced the classification model performance by 10% to 20% through the application of grid search and feature reengineering

    Key words: categorical variable, feature engineering, logistic regression, supervised learning

  • Wastewater Treatment Plant Design via Simulation and Optimization

    • Collaborated in a group of 4 to develop an optimization tool for Hatch Hydromantis to find the best possible wastewater treatment plant design using GPS-X
    • Developed a Python script based on the NSGA-II algorithm to optimize the WWTP design by minimizing economic costs and environmental impact
    • Awarded as the Best Industrial Application at the McMaster Engineering Capstone Expo Day

    Key words: wastewater, optimization, NSGA-II

  • Reinforcement Learning Based Controller Design Using Model Predictive Control

    • Researched offset-free approach for model predictive controller, actor-critic method, and issues with the DDPG algorithm
    • Solved continuous setpoints tracking issue in the Simulink reinforcement learning environment using Simulink SDI
    • Analyzed and compared the results of a continuous stirred-tank reactor model and a modified MATLAB water tank model

    Key words: CSTR, MPC, reinforcement learning, actor-critic, DDPG

  • Optimizing Supply Chain with Mixed Integer Linear Programming and Mixed Integer Nonlinear Programming Methods

    • Cooperated in a team of 3 to explore real-world optimization applications in supply chain management
    • Designed a dynamic supply chain network and formulated mixed integer linear/nonlinear programming models
    • Performed case studies on biodiesel supply and solved the problem using GAMS

    Key words: optimization, supply chain, MILP, MINLP, GAMS

  • Innovation in Resource Management

    • Joined a multi-disciplinary team of 6 to tackle the resource allocation problem of Aecon’s ESMSA team
    • Reviewed the existing status of ESMSA team using business model canvas and researched current market solutions to compare the solutions qualitatively and quantitatively in the integration and cost-benefit aspect
    • Maximized added value by proposing to extend the current construction management software license to incorporate the affiliated labour management software

    Key words: business model canvas, SWOT, PEST, resource management, cost-benefit analysis

  • Machine Learning Using Python

    • Initiated a self-directed learning project about various machine learning topics
    • Learned 20+ techniques of regression, classification, clustering, deep neural networks, natural language processing, dimensionality reduction, and decision making
    • Coded a stacked long short-term memory model making one-step and multi-step forecasting to solve the stock price prediction problem in the COMAP math modeling contest

    Key words: supervised learning, unsupervised learning, reinforcement learning, K-NN, K-Means++, SVM, decision tree, random forest, UCB, NLP, BoW, CNN, GAN, RNN, LSTM, SOM, RBM, SAE, XGBoost

  • Prediction of Rainfall in Melbourne, Australia

    • Collaborated in a group of 3 to work on Melbourne rainfall dataset to understand the correlations within variables and create a rainfall prediction model
    • Identified positive/negative relationships among variables and lowered the dimensions from 15 to 7 while keeping 90% of the original information using principal component analysis
    • Improved prediction accuracy by 48% using logistic regression to predict rainfall likelihood instead of using neural networks to predict the rainfall quantity

    Key words: PCA, PLS, ANN, logistic regression, supervised learning

  • Design of Experiments: Free Fall Time of Paper Airplane

    • Planned and conducted a DOE (26-2) to investigate major effect on paper airplane free fall time
    • Analyzed the experimental data and obtained factorial regression information using Minitab
    • Determined the next test set to increase the free fall time via the path steepest ascent method

    Key words: statistics, DOE

  • Chemical Process Synthesis and Simulation

    • Collaborated in a team of 4 to develop an innovative process that transforms waste plastics into ethylene and other valuable byproducts
    • Participated in designing a feasible chemical process and drawing a diagram using Lucidchart
    • Achieved 99% ethylene purity while emitting 78% less greenhouse gases than traditional process
    • Analyzed the life cycle inventory, supply chain, and the environmental impact using openLCA

    Key words: unit operations, chemical processing, sustainable energy