Skip to content
View huytjuh's full-sized avatar
  • Erasmus University Rotterdam
  • Amsterdam, NL
Block or Report

Block or report huytjuh

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
huytjuh/README.md

About Me

I am Data Scientist with a background in Econometrics and Operations Research. Specialized in Predictive Modeling, Causal Inference, and Optimization methods. Professional experience in all Machine Learning sub-disciplines, including NLP, Computer Vision, and Recommender Systems. Committed to consistency and transparency, dedicated to continuous improvements, and excited for new challenges.


Recent Projects πŸ’Ž

Vision-based Person Detection Model with YOLOv3 using Bayesian CycleGAN on Synthetic Data

An object detection model to predict person in indoor and hospital settings with a custom synthetically created dataset in Blender. An innovative solution to detect person instances without requiring actual person images by utilizing the state-of-the-art Bayesian CycleGAN to tackle down the challenging Synthetic-to-Real translation task.
Language: Python, Blender
Date: Q1 2022


Featured Projects πŸ†

Incorporating Subsequence Time-Series Clustering in LSTM, lightGBM, RF, Fourier-ARIMA, and Hybrid models on daily TV ratings of American TV-channels

Extending the state-of-the-art Time-Series forecasting models by utilizing a data-driven anomaly detection and complex seasonality clustering approach using Self-Organizing Map and Hierarchical Clustering methods. Results showed significant improvements on the existing forecasting models based on a 1-year ahead prediction of multiple time-series of over 10 TV-channels.
Language: Python
Date: Q1 2021

Open In Colab Open In Colab

Gaussian Mixture Model Clustering from scratch

Developed an advanced Gaussian Mixture Model from scratch by generating samples from a mixture of Gaussian distribution, improving the existing Machine Learning clustering methods and allowing for more complex clustering.
Language: Python
Date: Q4 2020

Open In Colab

General Framework of Constructing Hybrid Time-Series Models combining both Parametric and Machine Learning Models incl. RF, SVM, LSTM, (S)ARIMA

Created a general framework of how to construct hybrid models making full use of linear and non-linear models applied on a multiple time-series dataset. Key concepts such as outlier detection, correlation matrices and imputing missing datapoints are also discussed here.
Language: R
Date: Q3 2020

Open In Colab

Basket-sensitive Random Walk & Factorization Machine Recommendations applied on Grocery Shopping

Build a Recommender system based on Collaborative Filtering and extending it with a Facotrization Machine incl. ALS, RW. Suggested an alternative evaluation metric on measuring the performance of Recommender Systems in general.
Language: R
Date: Q2 2019

Open In Colab

Pinned

  1. YOLOv3-Blender YOLOv3-Blender Public

    YOLOv3 Blender implementation of privacy-preserving person-detection model using synthetic data.

    Python 2

  2. Domain-Adaptation Domain-Adaptation Public

  3. Subsequence-Time-Series-Clustering Subsequence-Time-Series-Clustering Public

    Extending state-of-the-art Time Series Forecasting with Subsequence Time Series (STS) Clustering to enforce model seasonality adaptation.

    Python 13

  4. Hybrid-Time-Series-Modeling Hybrid-Time-Series-Modeling Public

    Hybridization of Econometric and Machine Learning time-series models for cross-learning linear and non-linear patterns.

    Python 3 1

  5. Recommender-System-Basket-Analysis Recommender-System-Basket-Analysis Public

    Basket-Sensitive Recommender System & Factorization Machines for grocery shopping based on hybrid random walk models.

    R 2 1

  6. Adversarial-ML-Training Adversarial-ML-Training Public